15may27 Modularized framework for NEURON/Python network simulations with MPI

Overview

Using this modularized structure can select different cell types, populations, connectivities, etc. just by modifying the parameters in params.py.

An example is provided in params.py which switches from a single Hodgkin-Huxley population, randomly connected model (mpiHHTut); to a multiple Izhikevich populations, yfrac-dependent connected M1 model.

To test this just modify the variable simType in params.py: simType = ‘M1model’ # ‘mpiHHTut’

Bill’s description of what this repo should be

we should try to make this our main organizational code that can spin off multiple projects? – would like a clean code center-org so that not so painful as it has been with our prior code base (which btw was all my fault and not sam’s – he inherited from me) then any pushes back to this main center would have to be agreed by curator (you) and would generally involve just some small fix to a single file

README with description of the different files

List of files:

main.py: Main executable; calls functions from other modules.
params.py: Contains all static parameters. It is imported as “p” from all other modules, so that params can be referenced from any file using p.paramName
shared.py: Contains all the model shared variables and modules (except params.py). It is imported as “s” from all other file, so that any variable or module can be referenced from any file using s.varName
sim.py: Simulation control functions (eg. runSim)
network.py: Network related functions (eg. createCells)
cell.py: contains cell and population classes
conn.py: contains class ‘Conn’ used to instantiate connections, and contains methods to connect cells (eg. random, yfrac-based)
analysis.py: functions to plot and analyse data
stimuli.py: functions and parameters for differnt types of neural stimulation
izhi2007.mod: NMODL definition of Izhikevich 2007 neuron model
evol.py: Functions to run evolutionary algorithms (using inspyred) to optimize model parameters

Known issues, potential improvements, and alternative implementations (outdated - see TODO list at end of file)

- Saving only happens at the end (instead of at fixed intervals)

- Need to add a DEBUG mode or similar (like mpiHHTut had)

currently have a verbose flag, that prints some additional info
also have print messages along the code, which could be made optional

- Check structure of data saved to file

some nested lists with inhomogeneous dimensionality were converted to ‘object’ data types in order to be saved – check if structure makes sense when reading from python/matlab

- Some attributes/variables of the class ‘Cell’ and ‘Pop’ are not used but still initialized to [ ]

maybe can find alternative using different constructors for different cell types
fixed by using dictionary of tags - flexible for eahc type

- Labels (eg. AMPA, IT, HH,…) are defined by assigning numeric values to variables

can use dictionaries instead
need to check pros and cons of each – are dicts slower or have any limitations compared to lists/arrays ?
changed to using dictionaries

- stimuli.py not tested with this model

left there cause thought could be useful for future

- analysis.py still has many functions which havent been adapted to new structure

- Some variables inherited from cliff still have too long names, eg. backgroundSpikevec

- Lines too long, need to split using \

- if increase p.backgroundNoise > 0.0 get memory error

had to store rnadom noise generator separately for each cell

- Find a way to save lambda funcs (can’t be pickled)

can use: >>>import inspect >>>inspect.getsource(myfunc)

- Clearing vector in simdata now gives error because also has python variables

for v in s.simdata.itervalues() – fix error separated vectors

- ‘Cell’ class record method uses eval() which is unsafe

- currently setup to record traces from all cells

- bug when saving cell traces using multiple cores

Gathering spikes… >>> >>> >>> Traceback (most recent call last): File “main.py”, line 46, in <module> runSeq() File “main.py”, line 37, in runSeq s.sim.gatherData() File “sim.py”, line 117, in gatherData for d in gather: tmp.extend(d[k]) KeyError: ‘cellTraces_233’

now saved using different structure
had to remove hoc objects that are not pickable

- Load/save net+sim parameters from file

load net and sim parameters file from mat?
have different .py file to generate the params files

within params.py - have option to load from file, or set params and save to 2 files net and sim.mat

maybe functions in sim.py? can be called from params?

- Store net and sim params in dictionaries inside params.py

eg. p.net[‘ncells’]

facilitates saving to file (prev point)
can use if key in dict:
use regexp in submlime text to replace in all files:

Find What: p.(\w+) Replace With: p.sim[‘$1’]

- Use dict of tags/attributes for pop params and for cells

eg. p.net[‘popParams’][i][‘cellModel’]

can use if key in dict:

Define conn rules based on pair of tag/value for pre and post

eg. use tuple key: connProbs[‘name’,’IT’,’name’,’IT’] = (lambda x,y: 0.1*x+0.01/y) or dict of dict with tuple keys: connProbs[‘name’,’IT’][‘name’,’IT’] = (lambda x,y: 0.1*x+0.01/y)

- Add $Id$ hg info

can do with https://mercurial.selenic.com/wiki/KeywordExtension
but not recommended
bill wanted to do because cliff and I add last update and author info to files – which is never updated correctly
checked other github repos and they don’t have it - just eg. Contributors

- Rename main.py with init.py

Remove popType and connType variables - infer from dict keys

Remove popid from popParams - can use list index

Replace topClass and subClass with projection-type, type (neurotransmitter involved?)

email to Ben

Hey Ben, Im working on some of the changes we discussed. I’ve replaced variables with dictionaries of tags/attributes. For now, I’ve kep the ‘population’ concept, although can replace in future version if makes sense.

For both the ‘population’ and ‘cell’ objects you suggested replacing the ‘topClass’ and ‘subClass’ tags with ‘projectionType’ and ‘cellType’ if my notes are correct. I know projType for Exc cells will be ‘IT’, ‘PT’ or ‘CT’, but not sure what would be the best classification for Inh cells? Same thing for cellType, I think you mentioned neurotransmitters involved, but could you elaborate on what would be the list of possible values for both ‘Exc’ and ‘Inh’ cells/pops ?

We can use the google chat or this google doc to bounce ideas back and forth (link points to new section ready to be filled in).

Synapse

synapses as list of objects inside each cell (postsynaptic) - netcon in pre is stub; netcon in post is real synapse
netcon (neuron object) as part of synapse object

15dec28 Convert into python package

PyNet ? NeuPyNE, NeuPyNet, netpyne !! PYthon-based NETwork development framework for the NEuron simulator
make shared -> framework
from pynet import framework as f
just need init.py and param.py file
to add cells or conn functions use:

– class newCellClass(PointNeuron): … ; f.newCellClass = newCellClass – class newConnFunc(): … ; f.newConnFunc = newConnFunc

default simConfig
from pynet import init ; init.createAndRun(netParams, simConfig)

16jan26 Possibly making more simulator-indep using NeuroML-based format

NeuroML format

could represent all nml format using python dicts (instantiated net)
additional abstract layer (connParams, netParams etc) - using more general format and then converted to nml
eg. https://github.com/OpenSourceBrain/Thalamocortical/tree/master/neuroConstruct/generatedNeuroML2
https://www.neuroml.org/getneuroml
https://www.neuroml.org/tool_support
https://github.com/NeuralEnsemble/libNeuroML
https://github.com/NeuroML/pyNeuroML
http://bioportal.bioontology.org/ontologies/CNO
http://www.neuroconstruct.org/
http://neuronvisio.org/

Issues

Would like to keep simple declarative (python dicts based) specifications
Want to make netpyne more NEURON-independent, both at specifications, and py instantiated network
netpyne makes use of mod files, but neuroml requires detailed specification
netpyne provides conversion from abstract specifications -> instantiated network, and support for multicompartment, and subcellular connectivity
netpyne provides instantiation of NEURON objects, and parallel simulation of network
netpyne can use NeuroML structure but using dictionaries instead of xml text - so easy to manipulate
libneuroml and pynn - procedural, not so intuitive, and require internal definition of cell types etc (eg. IAF); whereas netpyne is ‘declarative’, more intuitive?, can define arbitrary cell props based on mechs and syns
dictionary of equivalences NEURON<->NeuroML (eg. tau1, rise)

cells

NOTE: segment = sections/segment; segmentGroup = sectionList/section

can have segmentGroups that consist of segmentGroups
Neuron sections->segmentGroups
Neuron 3d points -> segmnets
Neuron nseg -> property tag numberInternalDivisions
if have pointProcess in segment -> mapping of position along section and where pointprocess is placed
<include href=”nap.channel.nml”/>

<include href=”pas.channel.nml”/>

<notes>Cell: supaxax_0 exported from NEURON ModelView</notes>

connections

synapses

<?xml version=”1.0” encoding=”ISO-8859-1”?> <neuroml xmlns=”http://www.neuroml.org/schema/neuroml2” xmlns:xsi=”http://www.w3.org/2001/XMLSchema-instance” xsi:schemaLocation=”http://www.neuroml.org/schema/neuroml2 https://raw.github.com/NeuroML/NeuroML2/development/Schemas/NeuroML2/NeuroML_v2beta4.xsd” id=”Syn_AMPA_L4SS_IN”>

<notes>ChannelML file describing a single synaptic mechanism</notes>

<notes>Synapse with syn scaling constant c = 1 nS (translating to max cond of 2.94304e-07 mS), time course: 0.8 ms and reversal potential: 0 mV. Automatically generated by command: genSyn.py Syn_AMPA_L4SS_IN 0.8 1 0 </notes> </alphaSynapse>

</neuroml>

population

pulse generator (netstim?)

chat with Bill

summary of 3-hr meeting with Ben: 1) he really likes netpyne, 2) discussed how to add subcellular syn info (3 main methods: sectionlists, yfrac-based, path distance-based), 3) provided his unpublished code to convert sCRACM data to synapse density maps and said we can use in netpyne, 4) suggested making netpyne more NEURON/simulator-independent (similar to Kim Blackwell)

on this last point (simu-indep), I looked at NeuroML2 format, and some of the tools Padraig has built 2 interface it with Python and Neuron (eg. neuroconstruct, pyNeuroML, libNeuroML), and read Robert’s reproducibility paper

I think it might be possible to make netpyne simulator-independent at the specs and instantiated network, using a NeuroML-like internal format (but using python dicts instead of xml so can manipulate data) – and then keep final stage where its converted to Neuron simulatable objects

sent padraig an email to discuss the technical details of what they’ve already implemented and how netpyne could fit, and be useful to fill in missing tools in this whole story

an alternative is to keep it as a tool specific for Neuron, with its internal Neuron-based format for specs and network, and the possibility of exporting/importing from NeuroML and others (ie. its current status)

chat with Padraig

good to separate M1 model from netpyne
similarities with pnn and libneuroml
important of neuroml is conceptual structure
libneuroml has cell types
mapping from pyneuroml to mod files, but not the other way around
standardization of abstract conn rules is above pynn and nueroml/pyneuroml
eg 10 conn algorithms -> pynn -> neuroml
morphology - segments, segment groups
mod files would need to be translated into nueroml (not fully translated); alternative would be point to nueroml definition, convert via pyneuroml to Neuron mod
setup.py to netpyne
dot.travis.yml continuous integration server - run tests every time there is a commit to repository - make sure its consistent with latest version
convert from Neuron model to NeuroML:

– export specifications – OSB converting to NeuroML2 cell morphology (pyneuroml.convert_to_neuroml2) - in future if already in neuroml dont need to define in Nueron/netpyne format – Izhikevich (params) - abstract definiton – export to Nueroml2 positions and connectivity - no current tool to convert – https://github.com/OpenSourceBrain/Cerebellum3DDemo – http://www.opensourcebrain.org/docs#Converting_To_NeuroML2

convert from NeuromML to Neuron:

– java script to convert neuroml to mod and Neuron/py files - wait for better iteration – adapt java code to convert directly to netpyne – https://github.com/NeuroML/org.neuroml.export/blob/development/src/main/java/org/neuroml/export/neuron/NeuronWriter.java – https://github.com/NeuroML/org.neuroml.export/blob/development/src/main/java/org/neuroml/export/pynn/PyNNWriter.java – check param files hhtut izhitut updated so can padraig can modify – add load cell function in netpyne in neuroml format

differences:

declarative
netpyne makes use of mod files, but neuroml requires detailed specification
netpyne provides conversion from abstract specifications -> instantiated network, and support for multicompartment, and subcellular connectivity
netpyne provides instantiation of NEURON objects, and parallel simulation of network

options:

– long-term: use Neuroml internally for instantiated network – short-term: Neuron-based and conversion to NeuroML;

16jan27 Consolidate connectivity functions

Full conn -> fullConn

=probConn with prob = 1?

combine convConn, divConn, and randConn?

probFunc = optional param – if not present, then:
select function depending of param included? eg. ‘convergence’, ‘divergence’, ‘probability’

Random conn with N conns per postsyn cell (convergence) -> convConn

~= conv conn

conv param allowed as function eg. uniform, gauss
also allows for weight+delay as func

Random conn with N conns per presyn cell (divergence) -> divConn

Random conn with fixed, function (including distance-dep) prob, weight, delay -> probConn

fixed

number eg. 5

func

uniform

func dist-dep

function of yfrac
function of 3d distance
generalizable to func of

– (pre, post, dist) + ’’ + (xfrac, yfrac, zfrac, x, y, z) = 18 variables – (dist) + ’’ + (xznorm, xyznorm, xz, xyz) = 4 variables – total: 22 variables

create dictionary with each variable associated to its value/variable/function – don’t evaluate to save
enter as string, and convert to function
other variables allowed: any that are defined in f.net.params[var] – arbitrary

implementation

use **kwargs
string based - converted to function, extract variables
arguments weight and delay valid for all conn types; and can be functional

Options to implement func:

lambda with kwargs - NO

prob = lambda(**d): d[‘pre_ynorm’]*0.5

lambda with args - NO!

prob = lambda(*d): d[3]*0.5

string

prob = ‘pre_ynorm*0.5’ prob = ‘post_xyznorm*0.5 prob = ‘uniform()’ prob = ‘normal(10,5)’ prob = ‘uniform(1,5)’

functions allow all of python random functions
internal implementation:

– identifiy variables (pre, post or dist x,y,z,2d or 3d; plus any arbitrary defined in netParams) – define dict mapping variable name to lambda function returning actual variable using preCell and postCell as args – convert string to lambda function with variables as arguments – call lambda function with args as dict with each entry mapped to lambda func that returns the variable

efficiency

calculate all delay, prob, weight funcs together at the beginning
calcualte weights+delays for all possible conns despite some will not be used?
alternative might be more comput costly (calling 1 by 1), and repeated code
use map(lambda, list)
calculate final list of weights,probs,delays and pass to specific conn func

errors

arguments by reference using lambda func only works for ‘delay’ if ‘convergence’ func loops over preCellsTags and postCells – even though they have nothing to do with each other!! whats going on? python bug? Im missing something?!

comparison to PyNN

class AllToAllConnector(allow_self_connections=True, safe=True, callback=None)[source]

=fullConn

class OneToOneConnector(safe=True, callback=None)[source]

=conn rule specifying cell ids

class FixedProbabilityConnector(p_connect, allow_self_connections=True, rng=None, safe=True, callback=None)[source]

=probConn

class FromListConnector(conn_list, column_names=None, safe=True, callback=None)[source]

=conn rule specifying cell ids

class FromFileConnector(file, distributed=False, safe=True, callback=None)[source]

=conn rule specifying cell ids

class ArrayConnector(array, safe=True, callback=None)[source]

=conn rule specifying cell ids

class FixedNumberPreConnector(n, allow_self_connections=True, with_replacement=False, rng=None, safe=True, callback=None)

=divConn

class DistanceDependentProbabilityConnector(d_expression, allow_self_connections=True, rng=None, safe=True, callback=None)[source]

= probConn with dist-dep func

class IndexBasedProbabilityConnector(index_expression, allow_self_connections=True, rng=None, safe=True, callback=None)[sour

= conn rule specifying cell ids

class SmallWorldConnector(degree, rewiring, allow_self_connections=True, n_connections=None, rng=None, safe=True, callback=N

not implemented

class CSAConnector(cset, safe=True, callback=None)[source]

not implemented

comparison to NEST

ConvergentConnect

Connect many source nodes to one target node. =convConn

DivergentConnect

Connect one source node to many target nodes. = divConn

RandomConvergentConnect

Randomly connect one source node to many target nodes. = convConn with uniform func

RandomDivergentConnect

Randomly connect one source node to many target nodes. = divConn with uniform func

BinomialConvergentConnect

Connect a target to a binomial number of sources. ??

16feb18 Changing folder structure and adding setup.py

file structure

root

– examples/: param and run files – netpyne/: actual package – doc/: documentation – setup.py: required for pip install – other root files

OK how to add $PYTHONPATH in setup.py?

OK Make sure all examples and tuts work

examples - ok
tutorial examples - ok
claus - ok
M1Network -ok

OK Update documentation

OK Update README

16feb25 Adding STDP and RL

old stdp code

if s.usestdp and ([s.cellpops[pregid],s.cellpops[pstgid]] in s.plastConns): # If using STDP and these pops are set to be plastic connections if sum(abs(s.stdprates[s.EorI[pregid],:]))>0 or sum(abs(s.RLrates[s.EorI[pregid],:]))>0: # Don’t create an STDP connection if the learning rates are zero for r in range(s.nreceptors): # Need a different STDP instances for each receptor if newcon.weight[r]>0: # Only make them for nonzero connections stdpmech = h.STDP(0,sec=s.dummies[pstid]) # Create STDP adjuster stdpmech.hebbwt = s.stdprates[s.EorI[pregid],0] # Potentiation rate stdpmech.antiwt = s.stdprates[s.EorI[pregid],1] # Depression rate stdpmech.wmax = s.maxweight # Maximum synaptic weight precon = s.pc.gid_connect(pregid,stdpmech); precon.weight[0] = 1 # Send presynaptic spikes to the STDP adjuster pstcon = s.pc.gid_connect(pstgid,stdpmech); pstcon.weight[0] = -1 # Send postsynaptic spikes to the STDP adjuster h.setpointer(s.connlist[-1]._ref_weight[r],’synweight’,stdpmech) # Associate the STDP adjuster with this weight s.stdpmechs.append(stdpmech) # Save STDP adjuster s.precons.append(precon) # Save presynaptic spike source s.pstcons.append(pstcon) # Save postsynaptic spike source s.stdpconndata.append([pregid,pstgid,r]) # Store presynaptic cell ID, postsynaptic, and receptor if s.useRL: # using RL stdpmech.RLon = 1 # make sure RL is on stdpmech.RLhebbwt = s.RLrates[s.EorI[pregid],0] # Potentiation rate stdpmech.RLantiwt = s.RLrates[s.EorI[pregid],1] # Depression rate stdpmech.tauhebb = stdpmech.tauanti = s.stdpwin # stdp time constant(ms) stdpmech.RLwindhebb = stdpmech.RLwindhebb = s.eligwin # RL eligibility trace window length (ms) stdpmech.useRLexp = s.useRLexp # RL stdpmech.softthresh = s.useRLsoft # RL soft-thresholding else: stdpmech.RLon = 0 # make sure RL is off

how to update at intervals

why needed?

periodic saves (eg. of weights)
update virtual arm apparatus
run RL on syns based on critic signal
calculate LFP

option 1: run sim for short periods of time (not efficient?)

while round(h.t) < s.duration: s.pc.psolve(min(s.duration,h.t+s.loopstep)

option 2: finitialize handler?

netpyne format

within conn rule
One set of STDP params for each conn rule:

STDPparams = {‘hebbwt’: , ‘antiwt’:, ‘wmax’:, ‘RLon’: , ‘RLhebbwt’, ‘RLantiwt’, ‘tauhebb’, ‘RLwindhebb’, ‘useRLexp’, ‘softthresh’

stdpmech.RLon = 1 # make sure RL is on stdpmech.RLhebbwt = s.RLrates[s.EorI[pregid],0] # Potentiation rate stdpmech.RLantiwt = s.RLrates[s.EorI[pregid],1] # Depression rate stdpmech.tauhebb = stdpmech.tauanti = s.stdpwin # stdp time constant(ms) stdpmech.RLwindhebb = stdpmech.RLwindhebb = s.eligwin # RL eligibility trace window length (ms) stdpmech.useRLexp = s.useRLexp # RL stdpmech.softthresh = s.useRLsoft # RL soft-thresholding ‘plasticity’: {‘mech’: ‘STDP’, ‘params’: STDPparams}

Single set of STDP params for all:

‘plasticity’: ‘STDP’

revisiting STDP mod implementation

stdp.m1ms

COMMENT

STDP + RL weight adjuster mechanism

Original STDP code adapted from: http://senselab.med.yale.edu/modeldb/showmodel.asp?model=64261&file=\bfstdp\stdwa_songabbott.mod

Adapted to implement a “nearest-neighbor spike-interaction” model (see Scholarpedia article on STDP) that just looks at the last-seen pre- and post-synaptic spikes, and implementing a reinforcement learning algorithm based on (Chadderdon et al., 2012): http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0047251

Example Python usage:

from neuron import h

## Create cells dummy = h.Section() # Create a dummy section to put the point processes in ncells = 2 cells = [] for c in range(ncells): cells.append(h.IntFire4(0,sec=dummy)) # Create the cells

## Create synapses threshold = 10 # Set voltage threshold delay = 1 # Set connection delay singlesyn = h.NetCon(cells[0],cells[1], threshold, delay, 0.5) # Create a connection between the cells stdpmech = h.STDP(0,sec=dummy) # Create the STDP mechanism presyn = h.NetCon(cells[0],stdpmech, threshold, delay, 1) # Feed presynaptic spikes to the STDP mechanism – must have weight >0 pstsyn = h.NetCon(cells[1],stdpmech, threshold, delay, -1) # Feed postsynaptic spikes to the STDP mechanism – must have weight <0 h.setpointer(singlesyn._ref_weight[0],’synweight’,stdpmech) # Point the STDP mechanism to the connection weight

Version: 2013oct24 by cliffk

ENDCOMMENT

NEURON { POINT_PROCESS STDP : Definition of mechanism POINTER synweight : Pointer to the weight (in a NetCon object) to be adjusted. RANGE tauhebb, tauanti : LTP/LTD decay time constants (in ms) for the Hebbian (pre-before-post-synaptic spikes), and anti-Hebbian (post-before-pre-synaptic) cases. RANGE hebbwt, antiwt : Maximal adjustment (can be positive or negative) for Hebbian and anti-Hebbian cases (i.e., as inter-spike interval approaches zero). This should be set positive for LTP and negative for LTD. RANGE RLwindhebb, RLwindanti : Maximum interval between pre- and post-synaptic events for an starting an eligibility trace. There are separate ones for the Hebbian and anti-Hebbian events. RANGE useRLexp : Use exponentially decaying eligibility traces? If 0, then the eligibility traces are binary, turning on at the beginning and completely off after time has passed corresponding to RLlen. RANGE RLlenhebb, RLlenanti : Length of the eligibility Hebbian and anti-Hebbian eligibility traces, or the decay time constants if the traces are decaying exponentials. RANGE RLhebbwt, RLantiwt : Maximum gains to be applied to the reward or punishing signal by Hebbian and anti-Hebbian eligibility traces. RANGE wmax : The maximum weight for the synapse. RANGE softthresh : Flag turning on “soft thresholding” for the maximal adjustment parameters. RANGE STDPon : Flag for turning STDP adjustment on / off. RANGE RLon : Flag for turning RL adjustment on / off. RANGE verbose : Flag for turning off prints of weight update events for debugging. RANGE tlastpre, tlastpost : Remembered times for last pre- and post-synaptic spikes. RANGE tlasthebbelig, tlastantielig : Remembered times for Hebbian anti-Hebbian eligibility traces. RANGE interval : Interval between current time t and previous spike. RANGE deltaw : The calculated weight change. RANGE newweight : New calculated weight. }

ASSIGNED { synweight tlastpre (ms) tlastpost (ms) tlasthebbelig (ms) tlastantielig (ms) interval (ms) deltaw newweight }

INITIAL { tlastpre = -1 : no spike yet tlastpost = -1 : no spike yet tlasthebbelig = -1 : no eligibility yet tlastantielig = -1 : no eligibility yet interval = 0 deltaw = 0 newweight = 0 }

PARAMETER { tauhebb = 10 (ms) tauanti = 10 (ms) hebbwt = 1.0 antiwt = -1.0 RLwindhebb = 10 (ms) RLwindanti = 10 (ms) useRLexp = 0 : default to using binary eligibility traces RLlenhebb = 100 (ms) RLlenanti = 100 (ms) RLhebbwt = 1.0 RLantiwt = -1.0 wmax = 15.0 softthresh = 0 STDPon = 1 RLon = 1 verbose = 1 }

NET_RECEIVE (w) { deltaw = 0.0 : Default the weight change to 0.

Hebbian weight update happens 1ms later to check for simultaneous spikes (otherwise bug when using mpi)
if (verbose > 0)  { printf("flag %f\n",flag) }

if ((flag == -1) && (tlastpre != t-1)) { w = 0 : should be set to 0 for anti-hebb to work, but then there is weird error deltaw = hebbwt * exp(-interval / tauhebb) : Use the Hebbian decay to set the Hebbian weight adjustment. if (softthresh == 1) { deltaw = softthreshold(deltaw) } : If we have soft-thresholding on, apply it. if (verbose > 0) { printf(“Hebbian STDP event: t = %f ms; tlastpre = %f; w = %f; deltaw = %f\n”,t,tlastpre,w,deltaw) } : Show weight update information if debugging on. }

Ant-hebbian weight update happens 1ms later to check for simultaneous spikes (otherwise bug when using mpi)

else if ((flag == 1) && (tlastpost != t-1)) { :update weight 1ms later to check for simultaneous spikes (otherwise bug when using mpi) w = 0 : should be set to 0 for anti-hebb to work, but then there is weird error deltaw = antiwt * exp(interval / tauanti) : Use the anti-Hebbian decay to set the anti-Hebbian weight adjustment. if (softthresh == 1) { deltaw = softthreshold(deltaw) } : If we have soft-thresholding on, apply it. if (verbose > 0) { printf(“anti-Hebbian STDP event: t = %f ms; deltaw = %f\n”,t,deltaw) } : Show weight update information if debugging on. }

If we receive a non-negative weight value, we are receiving a pre-synaptic spike (and thus need to check for an anti-Hebbian event, since the post-synaptic weight must be earlier).

if (w >= 0) { interval = tlastpost - t : Get the interval; interval is negative if ((tlastpost > -1) && (interval > 1.0)) { : If we had a post-synaptic spike and a non-zero interval… if (STDPon == 1) { : If STDP learning is turned on… if (verbose > 0) {printf(“net_send(1,1)\n”)} net_send(1,1) : instead of updating weight directly, use net_send to check if simultaneous spike occurred (otherwise bug when using mpi) } if ((RLon == 1) && (-interval <= RLwindanti)) { tlastantielig = t } : If RL and anti-Hebbian eligibility traces are turned on, and the interval falls within the maximum window for eligibility, remember the eligibilty trace start at the current time. } tlastpre = t : Remember the current spike time for next NET_RECEIVE.

Else, if we receive a negative weight value, we are receiving a post-synaptic spike (and thus need to check for a Hebbian event, since the pre-synaptic weight must be earlier).

} else { interval = t - tlastpre : Get the interval; interval is positive if ((tlastpre > -1) && (interval > 1.0)) { : If we had a pre-synaptic spike and a non-zero interval… if (STDPon == 1) { : If STDP learning is turned on… if (verbose > 0) {printf(“net_send(1,-1)\n”)} net_send(1,-1) : instead of updating weight directly, use net_send to check if simultaneous spike occurred (otherwise bug when using mpi) } if ((RLon == 1) && (interval <= RLwindhebb)) { tlasthebbelig = t} : If RL and Hebbian eligibility traces are turned on, and the interval falls within the maximum window for eligibility, remember the eligibilty trace start at the current time. } tlastpost = t : Remember the current spike time for next NET_RECEIVE. } adjustweight(deltaw) : Adjust the weight. }

PROCEDURE reward_punish(reinf) { if (RLon == 1) { : If RL is turned on… deltaw = 0.0 : Start the weight change as being 0. deltaw = deltaw + reinf * hebbRL() : If we have the Hebbian eligibility traces on, add their effect in. deltaw = deltaw + reinf * antiRL() : If we have the anti-Hebbian eligibility traces on, add their effect in. if (softthresh == 1) { deltaw = softthreshold(deltaw) } : If we have soft-thresholding on, apply it. adjustweight(deltaw) : Adjust the weight. if (verbose > 0) { printf(“RL event: t = %f ms; reinf = %f; RLhebbwt = %f; RLlenhebb = %f; tlasthebbelig = %f; deltaw = %f\n”,t,reinf,RLhebbwt,RLlenhebb,tlasthebbelig, deltaw) } : Show weight update information if debugging on. } }

FUNCTION hebbRL() { if ((RLon == 0) || (tlasthebbelig < 0.0)) { hebbRL = 0.0 } : If RL is turned off or eligibility has not occurred yet, return 0.0. else if (useRLexp == 0) { : If we are using a binary (i.e. square-wave) eligibility traces… if (t - tlasthebbelig <= RLlenhebb) { hebbRL = RLhebbwt } : If we are within the length of the eligibility trace… else { hebbRL = 0.0 } : Otherwise (outside the length), return 0.0. } else { hebbRL = RLhebbwt * exp((tlasthebbelig - t) / RLlenhebb) } : Otherwise (if we’re using an exponential decay traces)…use the Hebbian decay to calculate the gain.

}

FUNCTION antiRL() { if ((RLon == 0) || (tlastantielig < 0.0)) { antiRL = 0.0 } : If RL is turned off or eligibility has not occurred yet, return 0.0. else if (useRLexp == 0) { : If we are using a binary (i.e. square-wave) eligibility traces… if (t - tlastantielig <= RLlenanti) { antiRL = RLantiwt } : If we are within the length of the eligibility trace… else {antiRL = 0.0 } : Otherwise (outside the length), return 0.0. } else { antiRL = RLantiwt * exp((tlastantielig - t) / RLlenanti) } : Otherwise (if we’re using an exponential decay traces), use the anti-Hebbian decay to calculate the gain. }

FUNCTION softthreshold(rawwc) { if (rawwc >= 0) { softthreshold = rawwc * (1.0 - synweight / wmax) } : If the weight change is non-negative, scale by 1 - weight / wmax. else { softthreshold = rawwc * synweight / wmax } : Otherwise (the weight change is negative), scale by weight / wmax. }

PROCEDURE adjustweight(wc) { synweight = synweight + wc : apply the synaptic modification, and then clip the weight if necessary to make sure it’s between 0 and wmax. if (synweight > wmax) { synweight = wmax } if (synweight < 0) { synweight = 0 } }

stdp.cliff

COMMENT

STDP + RL weight adjuster mechanism

Original STDP code adapted from: http://senselab.med.yale.edu/modeldb/showmodel.asp?model=64261&file=\bfstdp\stdwa_songabbott.mod

Adapted to implement a “nearest-neighbor spike-interaction” model (see Scholarpedia article on STDP) that just looks at the last-seen pre- and post-synaptic spikes, and implementing a reinforcement learning algorithm based on (Chadderdon et al., 2012): http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0047251

Example Python usage:

from neuron import h

## Create cells dummy = h.Section() # Create a dummy section to put the point processes in ncells = 2 cells = [] for c in range(ncells): cells.append(h.IntFire4(0,sec=dummy)) # Create the cells

## Create synapses threshold = 10 # Set voltage threshold delay = 1 # Set connection delay singlesyn = h.NetCon(cells[0],cells[1], threshold, delay, 0.5) # Create a connection between the cells stdpmech = h.STDP(0,sec=dummy) # Create the STDP mechanism presyn = h.NetCon(cells[0],stdpmech, threshold, delay, 1) # Feed presynaptic spikes to the STDP mechanism – must have weight >0 pstsyn = h.NetCon(cells[1],stdpmech, threshold, delay, -1) # Feed postsynaptic spikes to the STDP mechanism – must have weight <0 h.setpointer(singlesyn._ref_weight[0],’synweight’,stdpmech) # Point the STDP mechanism to the connection weight

Version: 2013oct24 by cliffk

ENDCOMMENT

NEURON { POINT_PROCESS STDP : Definition of mechanism POINTER synweight : Pointer to the weight (in a NetCon object) to be adjusted. RANGE tauhebb, tauanti : LTP/LTD decay time constants (in ms) for the Hebbian (pre-before-post-synaptic spikes), and anti-Hebbian (post-before-pre-synaptic) cases. RANGE hebbwt, antiwt : Maximal adjustment (can be positive or negative) for Hebbian and anti-Hebbian cases (i.e., as inter-spike interval approaches zero). This should be set positive for LTP and negative for LTD. RANGE RLwindhebb, RLwindanti : Maximum interval between pre- and post-synaptic events for an starting an eligibility trace. There are separate ones for the Hebbian and anti-Hebbian events. RANGE useRLexp : Use exponentially decaying eligibility traces? If 0, then the eligibility traces are binary, turning on at the beginning and completely off after time has passed corresponding to RLlen. RANGE RLlenhebb, RLlenanti : Length of the eligibility Hebbian and anti-Hebbian eligibility traces, or the decay time constants if the traces are decaying exponentials. RANGE RLhebbwt, RLantiwt : Maximum gains to be applied to the reward or punishing signal by Hebbian and anti-Hebbian eligibility traces. RANGE wmax : The maximum weight for the synapse. RANGE softthresh : Flag turning on “soft thresholding” for the maximal adjustment parameters. RANGE STDPon : Flag for turning STDP adjustment on / off. RANGE RLon : Flag for turning RL adjustment on / off. RANGE verbose : Flag for turning off prints of weight update events for debugging. RANGE tlastpre, tlastpost : Remembered times for last pre- and post-synaptic spikes. RANGE tlasthebbelig, tlastantielig : Remembered times for Hebbian anti-Hebbian eligibility traces. RANGE interval : Interval between current time t and previous spike. RANGE deltaw : The calculated weight change. RANGE newweight : New calculated weight. }

ASSIGNED { synweight tlastpre (ms) tlastpost (ms) tlasthebbelig (ms) tlastantielig (ms) interval (ms) deltaw newweight }

INITIAL { tlastpre = -1 : no spike yet tlastpost = -1 : no spike yet tlasthebbelig = -1 : no eligibility yet tlastantielig = -1 : no eligibility yet interval = 0 deltaw = 0 newweight = 0 }

PARAMETER { tauhebb = 10 (ms) tauanti = 10 (ms) hebbwt = 1.0 antiwt = -1.0 RLwindhebb = 10 (ms) RLwindanti = 10 (ms) useRLexp = 0 : default to using binary eligibility traces RLlenhebb = 100 (ms) RLlenanti = 100 (ms) RLhebbwt = 1.0 RLantiwt = -1.0 wmax = 15.0 softthresh = 0 STDPon = 1 RLon = 1 verbose = 0 }

NET_RECEIVE (w) { deltaw = 0.0 : Default the weight change to 0.

Hebbian weight update happens 1ms later to check for simultaneous spikes (otherwise bug when using mpi)

if ((flag == -1) && (tlastpre != t-1)) { w = -2 deltaw = hebbwt * exp(-interval / tauhebb) : Use the Hebbian decay to set the Hebbian weight adjustment. if (softthresh == 1) { deltaw = softthreshold(deltaw) } : If we have soft-thresholding on, apply it. if (verbose > 0) { printf(“Hebbian STDP event: t = %f ms; deltaw = %f\n”,t,deltaw) } : Show weight update information if debugging on. }

Ant-hebbian weight update happens 1ms later to check for simultaneous spikes (otherwise bug when using mpi)

else if ((flag == 1) && (tlastpost != t-1)) { :update weight 1ms later to check for simultaneous spikes (otherwise bug when using mpi) w = -2 deltaw = antiwt * exp(interval / tauanti) : Use the anti-Hebbian decay to set the anti-Hebbian weight adjustment. if (softthresh == 1) { deltaw = softthreshold(deltaw) } : If we have soft-thresholding on, apply it. if (verbose > 0) { printf(“anti-Hebbian STDP event: t = %f ms; deltaw = %f\n”,t,deltaw) } : Show weight update information if debugging on. }

If we receive a non-negative weight value, we are receiving a pre-synaptic spike (and thus need to check for an anti-Hebbian event, since the post-synaptic weight must be earlier).

if (w >= 0) { interval = tlastpost - t : Get the interval; interval is negative if ((tlastpost > -1) && (interval > 1.0)) { : If we had a post-synaptic spike and a non-zero interval… if (STDPon == 1) { : If STDP learning is turned on… net_send(1,1) : instead of updating weight directly, use net_send to check if simultaneous spike occurred (otherwise bug when using mpi) } if ((RLon == 1) && (-interval <= RLwindanti)) { tlastantielig = t } : If RL and anti-Hebbian eligibility traces are turned on, and the interval falls within the maximum window for eligibility, remember the eligibilty trace start at the current time. } tlastpre = t : Remember the current spike time for next NET_RECEIVE.

Else, if we receive a negative weight value, we are receiving a post-synaptic spike (and thus need to check for an anti-Hebbian event, since the post-synaptic weight must be earlier).

} else { interval = t - tlastpre : Get the interval; interval is positive if ((tlastpre > -1) && (interval > 1.0)) { : If we had a pre-synaptic spike and a non-zero interval… if (STDPon == 1) { : If STDP learning is turned on… net_send(1,-1) : instead of updating weight directly, use net_send to check if simultaneous spike occurred (otherwise bug when using mpi) } if ((RLon == 1) && (interval <= RLwindhebb)) { tlasthebbelig = t } : If RL and Hebbian eligibility traces are turned on, and the interval falls within the maximum window for eligibility, remember the eligibilty trace start at the current time. } tlastpost = t : Remember the current spike time for next NET_RECEIVE. } adjustweight(deltaw) : Adjust the weight. if (verbose>=1) { printf(“netreceive(): t=%f ms; w=%f; flag=%f; tlastpre=%f; tlastpost=%f, interval=%f; RLwindhebb=%f; tlasthebbelig=%f; RLwindanti=%f; tlastantielig=%f;\n”, t, w, flag, tlastpre, tlastpost, interval, RLwindhebb, tlasthebbelig, RLwindanti, tlastantielig) } : Show all information }

PROCEDURE reward_punish(reinf) { if (RLon == 1) { : If RL is turned on… deltaw = 0.0 : Start the weight change as being 0. deltaw = deltaw + reinf * hebbRL() : If we have the Hebbian eligibility traces on, add their effect in. deltaw = deltaw + reinf * antiRL() : If we have the anti-Hebbian eligibility traces on, add their effect in. if (softthresh == 1) { deltaw = softthreshold(deltaw) } : If we have soft-thresholding on, apply it. adjustweight(deltaw) : Adjust the weight. if (verbose>=2) { printf(“reward_punish(): t=%f ms; delta=%f; w=%f\n”, t, deltaw, synweight) } : Show weight update information if debugging on. if (verbose>=3) { printf(“reward_punish(): t=%f ms; delta=%f; w=%f; reinf=%f; hebbRL=%f; antiRL=%f; softthresh=%i\n”, t, deltaw, synweight, reinf, hebbRL(), antiRL(), softthresh) } : Show all information } }

FUNCTION hebbRL() { if ((RLon == 0) || (tlasthebbelig < 0.0)) { : If RL is turned off or eligibility has not occurred yet, return 0.0. hebbRL = 0.0 if (verbose > 2) { printf(“hebbRL(): off; t=%f ms; hebbRL=%f; RLon=%f; tlasthebbelig=%f\n”, t, hebbRL, RLon, tlasthebbelig) } : Show all information } else if (useRLexp == 0) { : If we are using a binary (i.e. square-wave) eligibility traces… if (t - tlasthebbelig <= RLlenhebb) { hebbRL = RLhebbwt } : If we are within the length of the eligibility trace… else { hebbRL = 0.0 } : Otherwise (outside the length), return 0.0. if (verbose > 2) { printf(“hebbRL(): square; t=%f ms; hebbRL=%f; tlasthebbelig=%f; RLlenhebb=%f\n”, t, hebbRL, tlasthebbelig, RLlenhebb) } : Show all information } else { hebbRL = RLhebbwt * exp((tlasthebbelig - t) / RLlenhebb) } : Otherwise (if we’re using an exponential decay traces)…use the Hebbian decay to calculate the gain. if (verbose > 2) { printf(“hebbRL(): exp; t=%f ms; hebbRL=%f; RLhebbwt=%f; tlasthebbelig=%f; RLlenhebb=%f\n”, t, hebbRL, RLhebbwt, tlasthebbelig, RLlenhebb) } : Show all information }

FUNCTION antiRL() { if ((RLon == 0) || (tlastantielig < 0.0)) { : If RL is turned off or eligibility has not occurred yet, return 0.0. antiRL = 0.0 if (verbose>=3) { printf(“antiRL(): off; t=%f ms; antiRL=%f; RLon=%f; tlastantielig=%f\n”, t, antiRL, RLon, tlastantielig) } : Show all information } else if (useRLexp == 0) { : If we are using a binary (i.e. square-wave) eligibility traces… if (t - tlastantielig <= RLlenanti) { antiRL = RLantiwt } : If we are within the length of the eligibility trace… else {antiRL = 0.0 } : Otherwise (outside the length), return 0.0. if (verbose>=3) { printf(“antiRL(): square; t=%f ms; antiRL=%f; tlastantielig=%f; RLlenanti=%f\n”, t, antiRL, tlastantielig, RLlenanti) } : Show all information } else { antiRL = RLantiwt * exp((tlastantielig - t) / RLlenanti) } : Otherwise (if we’re using an exponential decay traces), use the anti-Hebbian decay to calculate the gain. if (verbose>=3) { printf(“antiRL(): exp; t=%f ms; antiRL=%f; RLantiwt=%f; tlastantielig=%f; RLlenanti=%f\n”, t, antiRL, RLantiwt, tlastantielig, RLlenanti) } : Show all information }

FUNCTION softthreshold(rawwc) { if (rawwc >= 0) { softthreshold = rawwc * (1.0 - synweight / wmax) } : If the weight change is non-negative, scale by 1 - weight / wmax. else { softthreshold = rawwc * synweight / wmax } : Otherwise (the weight change is negative), scale by weight / wmax. }

PROCEDURE adjustweight(wc) { synweight = synweight + wc : apply the synaptic modification, and then clip the weight if necessary to make sure it’s between 0 and wmax. if (synweight > wmax) { synweight = wmax } if (synweight < 0) { synweight = 0 } if (verbose>=3) { printf(“adjustweight(): t=%f ms; synweight=%f; wc=%f; wmax=%f\n”, t, synweight, wc, wmax) } : Show all information }

netpynesimplestdp.mod

if use stdp.m1ms:

– only hebb events at the beginning – RL not working cause tlasthebbelig doesn’t update

if use stdp.cliff:

– both stdp and RL happen - but correctly?? – only hebb - so maybe anti-hebb not working – if critic = -1

problem might be netpyne?

m1ms model

if use stdp.cliff:

– seems like too many hebb events? – RL works (with critic =1 or -1) - but might be wrong due to hebb events?

if use stdp.m1ms:

– Hebbian works (no anti-Hebbian?) – RL works (with critic =1 or -1)

found bug and fixed (skip instead of w)

use separate ‘skip’ variable to skip the 2nd set of conditions; instead of using the ‘w’ var, since otherwise ‘w’ gets fixed to that value for following events

need to fix so -1 critic RL also works - fixed

was due to soft threshold being on – reduces -1 effect if close to min weight; and reduces +1 effect if close to max weight

check antihebb learning

fixed bug: if ((tlastpost > -1) && (-interval > 1.0)) (interval had to be -1)

16mar04 Adding subcellular synaptic conn distribution

sectionList

need during runtim? for connectivity purposes?
option 1: can include section attribute named ‘sectionList’ and add name, but could take longer to set properties if have to search all
option 2: can include as dict in sectionLists in cellProp, and then add all sections inside of each sectionList dict; but would change the consistency of the generic implementation (cell.py etc)
option 3: can add additional dict sectionLitst in cellProp, simply including list of sections (need to return additional dic from util.py function)
best options seem 1 or 3
Need to create sectionList at runtime? useful?

– iterating over sectionList from neuron import h s = h.Section(name=’s’) a = h.Section(name=’a’) d = h.Section(name=’d’) sl1 = h.SectionList() sl1.append(sec=s) sl2 = h.SectionList() sl2.append(sec=a) sl2.append(sec=d) section_lists = h.List(‘SectionList’) print ‘There are %d SectionLists’ % section_lists.count() for i in xrange(int(section_lists.count())): print ‘SectionList %d has the following sections: %r’ % (i + 1, [str(sec.hname()) for sec in section_lists.o(i)])

— Note that this iterates over all SectionList objects, not just those associated with a particular cell. — Alternatively, itertools.chain.from_iterable — https://docs.python.org/2/library/itertools.html#itertools.chain.from_iterable

NeuroML implements as option 3:

Netpyne implementation

Option 3

cellRule = {‘label’: ‘PYRrule’, ‘conditions’: {‘cellType’: ‘PYR’}, ‘secs’: {}, ‘secLists’: {}} … cellRule[‘secLists’][‘all’] = [‘soma’, ‘dend’]

Can’t define properties of section lists in netpyne
Neuron section list object not created - just python version
Can import cell with sectionlists
Can use sectionlists in connectivity
Modified importCell so it reads sectionLists with original variable names; and modified input arguments so now pass cellRule as 1st arg – so can add ‘secs’ and ‘secLists’ to cellRule dict

TODO:

update tut_import.py - DONE
update documentation - DONE
update claus - DONE
update M1Network - DONE

Discussion with Ben

define subcellular location of synapses:

named sectionlists of sublists - apicatuft, obliques
radial distribution, fraction yfrac, normalized, cortical depth, where dendrites fall in layers (cx centered, not cell centered)
path distance from soma

– plus combination of above – 1st, 2nd order branches differences cannot be specified directly, but can create sectionlists eg. 1st 50um apical trunk no spines - subsets that are spiny vs non-spiny

2 METHODS: A) 2-step implementation: 1) unitary conns, 2) contacts (subcellular conn) B) axo-dendritic density overlap - you don’t know unitary without calculating subcellular – HBP approach - axonal density is function of proj from L2->L5 + dendritic density, even if cells not realistic, calculate weight+prob

if we imagine a hierarch struc where a ‘unitaryConn’ object contains a list of ‘contacts’ (~= Neuron Netcons) and ‘synapses’ (Neuron Synapses) this unitaryConn would have a specific pre and post-cell id but it may be the case, that the synapse is used as the target of many presyn cells (at least in Neuron terms and how we usually build models now) which means the ‘synapse’ would have to be outside of the ‘unitaryConn’ and just be referenced by name or whatever I’ll just point out that this doesn’t make all that much sense to me (despite being the way you do it)

2 levels of instantiation - abstract (no specific models, not simulatable, cellular); and specific (subcellular)
from 1st - every neuron has to be assigned and implementation; and every connection has to be implemented
thats what happens right now but abstract instantiation is not saved
2nd step consists of instantiating subcellular info - cell morphology + subcellular conn (could be separate)
1st pass) cellular instatiation step (includes cell+conns); 2nd pass) subcellular step; 3rd pass) choose simulator pick Neuron object for each cell
In 3rd level decide where to place h.Netcons and h.Synapses
conns = unitary conns - presyn, type (AMPA), weight (somatic response for 1 AP presyn cell ~0.2 mV EPSP), delay?
in connParams have subcellular conn info - but not added to cell objects, only once create simulatable individual cells, wire it up
just 3 levels : specifications -> cellular -> simulator instantiation
Neurons instantiation separate from python structure - linked by cell id
for each unitary connection, check cell model of presyn and postsyn and decide if use NetCons, Synapses etc

Current implementation

Cell (object)

– conns (list of dicts) — hNetcon — preGid — weight — threshold — sec — syn

– secs (dict of dicts) — geom — topol — mechs — pointps — syns (dict) ---- hSyn ---- _loc ---- _type ---- tau ---- e

Pros/Cons of making conns indep of cells

pros:

– easier to search for conns, do analysis etc – conceptually make more sense (Ben, Bill) – consistent with NueroML

cons:

– synapses separate from cell too? associate via gid? prev models and Neuron standard is syns inside cell, conns outside; harder to check syns – make sure conns for each cell in same node – seems easy since will happen naturally – easier to find conns of a given cell

Possible structure (with contacts inside each unitary conn):

net (object of class Network)

– cells (list of objects of class Cell) – conns (list of dicts) — preGid — postGid — contacts (list of dicts) ---- hNetcon ---- weight ---- threshold ---- sec ---- syn

Advantage is that would have equivalent top level connections indepdently of subcellular info (morphology).
Advantage: Need to know all unitary connections before can actually implement the subcellular details - distribution depends on total number of input conns
Disadvantage is that have more complex hierarchical structure, more difficult to search, if have simple single contacts then overcomplicated, and doesnt match neuroml

Alternative structure (with single list of contacts):

net (object of class Network)

– cells (list of objects of class Cell) – conns (list of dicts) — preGid — postGid — hNetcon — weight — threshold — sec — syn

for unitary conns - find unique preGid/postGid (eg. print list(set(zip(*myList)[1])))

Alternative structure (with contacts inside, but as a class)

net (object of class Network)

– cells (list of objects of class Cell) – conns (list of object of class Connection) — preGid — postGid — contacts (list of dicts) ---- hNetcon ---- weight ---- threshold ---- sec ---- syn

Advantage - can have subcelullar methods inside

Structure with proper nomenclature for cellConns, synConns, synMechs

net (object of class Network)

– cells (list of objects of class Cell) – cellConns (list of object of class CellConnection) — preGid — postGid — synConns (list of dicts) ---- hNetcon ---- weight ---- threshold ---- sec ---- synMech

Choice of cellConns, synConns, synMechs

‘cellConns’ = unitary connection between 2 cells (can contain many synapses) = non-existent (Neuron) = non-existent (NeuroML)

= ‘connection’ (HBP)

‘synConns’: individual synaptic contacts made between single pre and post cell (each connection can contain many) = ‘h.NetCon’ (Neuron) = ‘connections’ (NeuroML) = ‘synapse’ (HBP)
‘synMechs’: postsynaptic mechanism (receptor-like), usually one per connection = eg. ‘h.ExpSyn’ or ‘h.AMPA’ (Neuron) = eg. ‘alphaSynapse’ (NeuroML) = ‘synaptic mechanism’ (HBP)

neuroml conn structure

projections: between populations; pre, post and synapse
connection: pregid, postgid, presegment, postsegment, prefraction, postfraction
synapse: mechanism associated to cell; tau, gbase, erev

<notes>ChannelML file describing a single synaptic mechanism</notes>

<notes>Synapse with syn scaling constant c = 1 nS (translating to max cond of 2.94304e-07 mS), time course: 0.8 ms and reversal potential: 0 mV. Automatically generated by command: genSyn.py Syn_AMPA_L4SS_IN 0.8 1 0 </notes> </alphaSynapse>

Ben’s sCRACM code

insertSyn

vecstim
netcon
expsyn

uniformSynapseLocations

total length
list of secs
sec lens
find total length, divide by num syns, find absolute locs, find local secs and locs for each syn (use list with cum lengths and corresponding section)

pointFromLoc

section, loc
return anatomical point x,y,z

loadratio

load file wiht synaptic dist ratios - rel syn strength for 10x30 grid

setup_grid

create grid with 10 cols, 30 rows

calculateGridSigma

vector of local synaptic strength at each grid point, e.g. ratio peak_experimental / peak_uniform

distanceFromGrid

// arg 1: x // arg 2: y // arg 3: Vector to hold the distances

interpolateSegmentSigma

arg 1: SectionList to receive synapses, e.g. spiny
arg 2: Vector to be filled with the calculated sigma
Note: the dendrites must be fully enclosed by the grid (i.e. there must be 4 nearest grid points surrounding every segment)
for each section loc:

– find x,y pos – find 4 closest grid points and their sigmas (syn strengths) – bilinear interpolation of sigma

mappedSynapseLocations

arg 1: SectionList for which to calculate sigma
arg 2: Vector of sigmas for each segment
arg 3: SectionList to be filled
arg 4: Vector of x-vals to be filled
Load ratio (vals between 0-15) of syn strength in 10x30 grid
Obtain norm sigma values based on ratio
Calculate sigma via interpolation for every segment location
Calculate synapses to insert in each segment

radialSynapseLocations

arg 1: SectionList that can accept synapses (e.g. spiny)
arg 2: sigma_uniform (e.g. 1.5/8)
arg 3: y_min
arg 4: y_max
arg 5: SectionList to carry synapses, 1 for each synapse (e.g. synsecs)
arg 6: Vector of x-values, 1 for each synapse (e.g. synxs)
uniform distibution of synapse based on radial distance from soma
specify y range
calculate num synapses based on section length / nseg
store section and loc of each syn

distributeSynapse

arg 1: SectionList that can accept synapses (e.g. spiny)
arg 2: sigma_uniform (e.g. 1.5/8)
arg 3: mode: 0 = uniform, 1 = non-uniform aka mapped, 2 = non-uniform based on laminar depth
arg 4: SectionList to carry synapses, 1 for each synapse (e.g. synsecs)
arg 5: Vector of x-values, 1 for each synapse (e.g. synxs)
arg 6: optional, for mode 2: y_min
arg 7: optional, for mode 2: y_max
wrapper: selects 1 of 3 modes and passes in arguments

netpyne implementation

check number of presyn cells connecting to post (cellCons)
check syn distribution pattern and morphology of post
need to calculate num of synConns based on weight?? or fixed weight for synConn (divide cellConn weight by num of synConns?)
create synConns and synMechs
Is it worth having CellConn as class with SynConn attribute and methods?

– only 5 syns/cell – need to know all cellconns before start synconns, but method cannot be inside cellconn cause need to check all cellconns – have to call method createSubCellConn if params includes subCell - do once for each conn rule (from network.py) – subcellular syn dist shouldn’t be replicated on each CellConn, just once on conn rule

createConns

– for each rule: — probConn/convConn ---- cellConns.append(CellConn) (but no subcellular) – for all cellConns: — cellConns.addSubcellular

but needs subcell params from rule!! so either

– need to be stored in each cellConn / pointer to f.net.params – shouldn’t occupy extra space since pointer (can remove after so not saved?)

do we need subcellular info for each conn rule?! or just for each cell type and synMech?! Ben mentioned ‘after logic conns, create subcellular based on presyn+postsyn cell type”
weight + delay a property of CellConn or SynConn ?! both?
how to create extra synMechs needed? identical properties to cellprop syn mechs?
maybe there’s no need for cellconns vs synconns?

– if subcellular info: create temporal cellconns, then synconns (from synconns can generate cellconns after) – if no subcellular info: create synconns directly

or create cellconns and synconns as independent lists (redundant)
maybe create separate rules for subcellular distribution; needed because:

– conn data as func of yfrac (eg. [0.1 - 0.2]) – syn dist for presyn layer/type (eg. L2/3) – wouldn’t make sense to repeat for each yfrac-based conn rule (or cell type rule), plus need all

maybe add subcellular distribution to cell prop rules:

– for each cell prop, create conditions of subcellular distribution eg. [‘conditions’: {‘yfrac’: [0.1-0.4]}] – default: assume uniform distribution

where to put properties of synapses?

– can add to sections in properties – can just add generic, by setting ‘_loc’: ‘generic’ ? – but still inside a section! :/ – just make algorithm look for synmechs in all sections, and use same params to replicate – maybe move ‘synMechs’ out, to same level as ‘secs’ ? – add to sectionList !?

sCRACM (Petreanu 2009; Suter,2015)

inject ChR2 in certain presyn axons (eg. L2/3)
short pohotostimulation 1ms 2mW in specific grid (12x24, 50um) location
depolarizes axons; use TTX (blocked Na+ and K+ ->no AP)
axon terminal Ca2+ channels open -> glutamate release
measure EPSC in soma
provides a two-dimensional ‘image’ of the distribution of specific input within the dendritic arbors of the recorded cell
EPSCsCRACM amplitudes depend on the density of ChR2-positive axons, the fraction of axons that make synapses with the recorded neuron, the strength of the synapses, and their electrotonic distance from the soma
sCRACM maps were normalized to the largest pixels within a map and thus represent the relative strength of input within the dendritic tree.
Average sCRACM map of layer 2/3 inputs to M1-CSPs (n = 23 neurons). L, Input profiles by soma depth. M, Location of perisomatic input from layer 2/3 relative to soma position, plotted as a function of soma depth. For each profile, the input depth was calculated as the center of mass across the perisomatic pixels.

fig

questions for ben

always have to specify num contacts or distribution?

– Yes, or distribution – Ben: if full model the could calculate exactly; but since scaled down, we want syns per conn to be correct eg. 5 – assume in M1 model, realistic density, but reduced volume – experiments found ~5 for PT->PT not in motor cx – likely that there is data – does it vary significantly? Thalamocortical could be very different; assume every proj has 4-6 contacts as measured – select a fixed value eg 5. and value strength/weight of each synConn

Weight+delay in cellConn or synConn or both?

– if data from unitary conn EPSP, the cellConn, but need to calculate weight/synConn (h.NetCon) – allow weights both at cellConn and synConn level – if weight * 2 makes the synapse twice as strong? – can know max conductance, and temp response – Lefort used EPSP mV in soma for cellConn – need to know how to calculate synConn – is there a way to do this analytically? – he did in scracm code- assume syn density is correct, electrotonic difference made different between measured and simualated – multiply by factor difference, resimulate and compare again – syns close to soma were fine; syns far away, original map had understimated response of far away syns, and rescaled far away syns – pattern of syn density for cell – function of yfrac – only need to calculate for original scracm map, and output should be rule – but for cellConn -> 5 synConns get 1mV in soma when activated simultaneously – but will be distributed along dendrites so different transmission delays – difficult to calculate accurately – most models dont distribute syns over dendrite with realistic syn density

– delay could be identical for each synConn, or add small distance-based delay – delay would vary due to extra distance traveled through unmyelinated axon (1 m/s) – 1.1ms diff for different pyramidal neurons - can omit for now - euclidean distance between cell bodies - condVelocity = 1m/s – minDelay value = 2ms – too high! set to maybe 0.5 ms or 1ms – max(0.5, 3d distance based) ; for closeby cells, axon goes down ~100um (fat axon), collaterals branch off, come back up 100 or 200um ; localDelay = 0.5ms – for long-range conns use longer delays - eg. 4 ms - kind of aritraty seems noise

Bill: have to normalize by position based on a transfer function between locations and soma

– can use someting like ted’s multifreq transfer func program to get tha map – yes have to scale baesd on loc and based on freq of input – eg a high freq (short) input like AMPA has less effect on soma than a slow eg NMDA – was looking at paper on that recently but is just for passive case – jnsci15:1669.pdf The morphoelectrotonic transform: a graphical approach to dendritic function /u/billl/articles/jnsci15:1669.pdf – reshaping tree based on electrical distance – another old paper – author = “Jaffe, DB and Carnevale, NT”, title = “Passive normalization of synaptic integration influenced by dendritic architecture”, /u/billl/articles/jnphys82:3268.pdf – http:/www.neuron.yale.edu/phpbb/viewtopic.php?f=8&t=2404&p=9487&hilit=frequency+response#p9487 – calculate the space constant in response to a AC current // lamda_AC(f) = 1/2*sqrt(diam/PI*f*Ra*Cm) – think he has a little package for this but not finding now http://www.neuron.yale.edu/phpbb/viewtopic.php?f=15&t=313&p=870&hilit=frequency+response#p870 a ‘frequency tool’ – also depends on the Rm and on active conductances, eg an EPSP can end up being terminated by K chans or can be primarily and active response so that the actual gsyn is only a small piece

Syn dist specific to postsyn mech? depends on presyn cell props,loc?

– Specific to eg. AMPA/NMDA, or just exc vs inh? – we dont have specific data on differnt receptors – assume all exc are both AMPA and NMDA, and have a conductance ratio between them - has been measured (maybe ratio different in tuft vs basal, but ignore for now); allow to customize ratio

depends on presyn cell type and layer (eg. L2/3 exc vs L5B IT vs L5B PT)

– M1 L2/3 -> SPI sCRACM map – Petreanu barrel cortex L2/3 or L4 -> yfrac sCRACM map – long range inputs sCRACM maps (both Ben and Petreanu)

Why need to know num of conns before start? Can have prob dist of synapse, and map on a cell-by-cell basis, adapt

probabilities based on current num contacts – wouldn’t be exact since probability based; but need final distribution to match ratios

The non-uniform mapped / distance from soma-based would be given in the form of 30x10 map, distribution across segments?

– exp data is in 30x10 map format; distr across segments would vary depending on cell morphology – netpyne should be a 30x10 length density – num of syns/um – within given pixel stimulated with blue light, pixel = 50x50 um, num of syns/um/dendrite - homogeneous within 50x50um – convert scracm map -> map of how many syns per length of dendrite at each region of the map (at each pixel) – need to know how much dendrite in that pixel, and how many sysn were coactivated to get soma response – his method: put fixed number of [of syns in] dendrite, and then adjusted weight of these syns – couldnt really separate weights from num of synapses – started with plausible weight and num of syns – length density for a specific assumption of conductances (used to convert map) – input to netpyne should be free from Neuron peculiarities – relates to somehting experimentally – num of syn per length of dendrite and strength of each of thsoe syns as a peak conductance

2 types of map: 1) num syns/um of dendrite, 2) conductance of syns/um dendrite (probably uniform everywhere)
remain within bounds that exp plausible - eg. now num of spines in a dend tree (eg max 20k)
L5->L5 somatic epsp unitary conns 0.2 0.5 mv for 5 syns/conn –> each syn has g that will result in 0.1 mV epsp soma

– but will depend on location eg apical tuft won’t give you value – 5 number comes mostly from counting dyns in basal – but maybe also in apical tuft but weren’t undetected (90% came from basals) – so probably ok to assume most syns are in basal not apical tuft – in his model he calculated conduc value for standard synapse that gives 0.1 or 0.2 mV (assume uniform over all arbor), so need to vary num of contacts and location

Other formats apart from length density map and conductance map (50um and 10x30 map,aligned to pia, centerd on soma):

– for STR assume they get same input and infer similar density map for STR cells – convert using dendritic density of STR cells – simpler way: syn length density maps are also radial functions – if account for how much dendrite there is, the x dimension of map may not contain any info – so rule is based on 1d map (radial function) - 30x1 map - function of yfrac (in 50 um bins) – could be averaged or represent individual neurons – for scracm map doesnt make sense to average – sectionLists - scracm has 50um resoulution but actually more like 100um resoulution – if you look at dends, 1st 50um of apical trunk have no syns; 1st 20um of root basal dont have any syns (just few) – at a scale where scracm doesnt reach there might be no syns – so where scracm doesnt reach can use sectionList to complement and tell u there are no syns thre – eg. spiny sectionList contains all sections except where they were no spines/syns - so set density to 0 – in neuron if u didnt reconstruct eveyr spine, you account for that by increasing cm – in detailed model I’m using – compensated by increasing membrane area

syn density maps are not common - so need other way to specify for rest – more common data in literature, anatomical/categorical statements eg. this interneuron targets apical tuft – but no specific numeric values (details not known) – usually always anatomical
he can send SPI syn length density map – SEND EMAIL!

comments by ben

If you want to see which dendrites are on basal, apical, etc. then better to look at the profiles sorted by soma depth, i.e. where you can see each individual neuron’s input collapsed to a radial profile.

[5:17] Or a soma-aligned average map.

[5:18] The former is in my paper. The latter I have as figs.

[5:19] In my paper I also show analysis of where the “perisomatic” input is centered, relative to the soma - generally above the soma with an offset of ~50 um, but depends on projection and on soma depth.

[5:19] One consistent finding was that while the basals do get input from all sources, it’s a bit weaker below the soma (note that there are basals above the soma too).

[5:21] The sCRACM technique probably won’t work well for local L5->L5 connections (Taro tried this with Rabies - not mapping, but LED flash - and said the input was too weak). But for L5-L5 connections we can turn to pair recording studies with full reconstructions of dendrites and axons - such as published by Markram et al 1997 (methinks), with Sakmann.

There is substantial variability within each projection - seen clearly in the example Salva just pasted. Some CSP get strong L2/3 input, while others don’t - even if their soma depth is similar.

[5:29] So my thought was to take each actual map and apply the inferred synapse density onto reconstructions at approximately matched soma depths. Do this for each projection.

[5:29] For the L2/3 -> CSP data set, I have reconstructions and maps in the same neuron, for a subset of the data set. For the other projections, I don’t have reconstructions.

[5:30] So the idea would be to take each mapped neuron, use it’s soma depth to find the reconstruction with closest matching soma depth, and instantiate that.

[5:31] Or alternatively (and probably better), take each reconstructed neuron, and then apply synapses to it according to every projection - where for each projection, you find the map whose soma depth best matches the reconstructed neuron.

[5:32] Another approach I started working on is to parameterize each projection’s pattern as a function of soma depth, so that there is a formula M_i(somadepth) for each projection i, which lets you determine the synapse distribution for neuron’s at any depth.

[5:34] However, because the sCRACM maps don’t tell you synapse density (but rather, synaptic strength measured at soma under strange pharmacological conditions), before you can calculate a formula for the whole projection, you’d need to simulate each map to infer the synapse density. Unless … it turns out that there is a nice analytic way to convert the input maps into synapse densities, i.e. without simulating each one. From the few simulations I’ve done, this might actually be possible.

Features implemented with adhoc solution – need to be fixed

GENERAL

default synMech placed in soma - where to define? make syns separate like neuroml

don’t create synMechs until needed by connections – just specify params? but need to store somewhere
can look up in f.net.params using tags[‘propList’] but probably slow
in Cell class have pointer to f.net.params from synMechs but dont create new python/neuron object
place synapse together with connection instead of cell? confusing since many conns can project to single synapse
In neuroml, syns are defined independently, and associated with connections - do same thing:

netParams[‘synMechParams’] = {‘NMDA’: {‘mod’: ‘ExpSyn’, ‘tau’: 0.1, ‘e’: 0}} cell.secs[‘soma’][‘synMechs’][0][‘label’: ‘NMDA’, ‘loc’: 0.5, ‘hSyn’: ]

place synMechs outside of sec? might be slower to find; plus synMechs are pointps so ‘belong’ to section
no need to add in netParams, specified in connRule, and created together with connections
when create conn check if synMech exists in that section, ie. has same ‘label’ and ‘loc’, if not, then create
make addSyn function
comparison of searching inside list vs accessing dict (~20x faster, for 1M conns: 0.75 sec vs 0.03sec; for 100M conns: 75 sec vs 4 sec):

– but have to add time of searching for specific location too

timeit.timeit(“b=next((item for item in a if item[‘label’]==’ampa’), None)”, setup=”a=[{‘label’:’gabab’,’v’:10},{‘label’:’gaba’,’v’:10},{‘label’:’nmda’,’v’:10},{‘label’:’ampa’,’v’:10}]”, number=1000000) Out[13]: 0.7533509731292725

timeit.timeit(“b=a[‘ampa’]”, setup=”a={‘gabab’:{‘v’:10},’gaba’:{‘v’:10},’nmda’:{‘v’:10},’ampa’:{‘v’:10}}”, number=1000000) Out[18]: 0.038023948669433594

NEED TO IMPORT SYNAPSES INTO SYNMECHPARAMS

independent scale conn weight for each cell model

where to store?
useful feature
maybe by default should be stored by cell model
allow both options (same variable)

move conns and stims out of Cell class

make stims be normal populations?
stims from S2 projection could make 2000 syns into single cell!
maybe place netstim part of stim outside of cell, but keep netcon inside
maybe keep conns inside cell for now? seems to make things easier eg. have all synConns of that cell ready to distribute

CONN RULE

conn rule with list of synMechs

eg. [AMPA, NMDA]
weightFraction or weightRatio - eg.
synWeightFraction = [0.9, 0.1]

synsPerConn as distribution

allows functional - seems to be gauss around 5 (E->) or 1 (I->)

weight/delay of conn and syns

allow to provide single weight – find way to estimate weight of syn contacts – check 16mar24 entry with papers to read!
allow to provide list of synweights – one per synConn (eg. 5 values) by convention assign starting from pia
allow functional only for single value (otherwise code would get overcomplicated)
maybe allow argument that represents location of synapse (not of cell) ??

SUBCELL RULE

create 1st conns, then syns

2 steps, so can distribute syns
create temporary cellConns or keep?
need somewhere to store unitary weight? or when convert to subcellular lose?
1) create synConns but don’t assign locs/delays/weights; 2) for each cell distribute syns, based on sep subcellular rules

– maybe add subcellular distribution to cell prop rules: — for each cell prop, create conditions of subcellular distribution eg. [‘conditions’: {‘yfrac’: [0.1-0.4]}] — default: assume uniform distribution — wouldn’t make sense to repeat for each yfrac-based conn rule, plus need all

create num synConns in 1st pass, and then replace if needed based on subcelluar rules
only do 2nd pass if any subcellular rules (can’t depend of numSyns cause could be 1, but still need to distribute)

distribute syns in sectionList or list of sectionLists

add suport for list of sectionLists - or even combination?!

syn distribution specification

2d syn density map
1d syn density map (yfrac)
apply 1d or 2d map only to certain sectionList, eg. spiny
categorical data - eg. by sectionLists + distribution function (uniform, gauss)

Meeting with Ben (March 18)

Comments from Ben

conceptual framework makes sense
HBP approach - axodendritic approach vs. know unitary and subcellular = refinement
2 netcons to same synapse - what happens if input at same timestep ? coincident spiking
1st pass + 2nd pass - approves
network model indep of subcellular info
single synapse could represent a projection between
unitary conn from preCell A could project to postCell B and C (so not totally accurate in terms of anatomy)
Do we need subcellular map for each presyn and postyn cell type, or per conn rule

– only pero conn rule if you average maps; but could also use indiv cell maps for each cell model – upper and deeper SPI cells scracm maps suggests they would have different input-output relations in networks – for a given density map could end with different syn density (and input-output relation) by applying to diff morphologies – yfrac dep of subcellular conn – also variabilty – CSP morphology very consistent for diff cortical depths; difference is in linker segments (eg 8% vs 12%) so not big diff – CSTR differences based on yfrac - within L5A and L5B - dont know if intrinsic diffs in CSTR

Make 3 CSP models - upper, middle, lower - 100 um apart – get data from 10 cells for each and fit – then test if can replace each of them and still get consistent results or very different (1 membrane model into 3 diff morphologies)
For IT have upper layer, L5A, L5B, and L6 – CSTR give you L5A, L5B, and have data for IT upper layers
IT cells - corticortical (corticocollosal; not exclusive) in L2/3; corticostriatal mostly in L5 (also corticocortical; not exclusive); not all 1 class (eg. project to S2 and M1, non-overlapping, intrinsics are different)
Naoki looked at IT L/3, L4 and L5 – different intrinsic? no consistent answer
cellConns vs synConns needed?

– single S2 prjection can make 2000-3000 synapses /contacts into single cell

Estimating synConn weights from cellConn weight

Discussion with Ben+Bill

Weight+delay in cellConn or synConn or both? (Ben)

– if data from unitary conn EPSP, the cellConn, but need to calculate weight/synConn (h.NetCon) – allow weights both at cellConn and synConn level – if weight * 2 makes the synapse twice as strong? – can know max conductance, and temp response – Lefort used EPSP mV in soma for cellConn – need to know how to calculate synConn – is there a way to do this analytically? – he did in scracm code- assume syn density is correct, electrotonic difference made different between measured and simualated – multiply by factor difference, resimulate and compare again – syns close to soma were fine; syns far away, original map had understimated response of far away syns, and rescaled far away syns – pattern of syn density for cell – function of yfrac – only need to calculate for original scracm map, and output should be rule – but for cellConn -> 5 synConns get 1mV in soma when activated simultaneously – but will be distributed along dendrites so different transmission delays – difficult to calculate accurately – most models dont distribute syns over dendrite with realistic syn density

– delay could be identical for each synConn, or add small distance-based delay – delay would vary due to extra distance traveled through unmyelinated axon (1 m/s) – 1.1ms diff for different pyramidal neurons - can omit for now - euclidean distance between cell bodies - condVelocity = 1m/s – minDelay value = 2ms – too high! set to maybe 0.5 ms or 1ms – max(0.5, 3d distance based) ; for closeby cells, axon goes down ~100um (fat axon), collaterals branch off, come back up 100 or 200um ; localDelay = 0.5ms – for long-range conns use longer delays - eg. 4 ms - kind of aritraty seems noise

Bill

Bill: have to normalize by position based on a transfer function between locations and soma

– can use someting like ted’s multifreq transfer func program to get tha map – yes have to scale baesd on loc and based on freq of input – eg a high freq (short) input like AMPA has less effect on soma than a slow eg NMDA – was looking at paper on that recently but is just for passive case – jnsci15:1669.pdf The morphoelectrotonic transform: a graphical approach to dendritic function /u/billl/articles/jnsci15:1669.pdf – reshaping tree based on electrical distance – another old paper – author = “Jaffe, DB and Carnevale, NT”, title = “Passive normalization of synaptic integration influenced by dendritic architecture”, /u/billl/articles/jnphys82:3268.pdf – http:/www.neuron.yale.edu/phpbb/viewtopic.php?f=8&t=2404&p=9487&hilit=frequency+response#p9487 – calculate the space constant in response to a AC current // lamda_AC(f) = 1/2*sqrt(diam/PI*f*Ra*Cm) – think he has a little package for this but not finding now http://www.neuron.yale.edu/phpbb/viewtopic.php?f=15&t=313&p=870&hilit=frequency+response#p870 a ‘frequency tool’ – also depends on the Rm and on active conductances, eg an EPSP can end up being terminated by K chans or can be primarily and active response so that the actual gsyn is only a small piece

TO READ or check out!

Mark15

-Pathway specific values for the parameter “utilization of synaptic efficacy” (U, analogous to the probability of neurotransmitter release) were unified from various experimental studies of synaptic transmission in juvenile rat somatosensory cortex (Le Bé et al., 2007; Brémaud et al., 2007; Feldmeyer, 2006; Koester and Johnston, 2005; Markram et al., 1997; Silver et al., 2003).

The axonal conduction delay for each synaptic contact was computed using the axonal path distance from the soma, and a AP conduction velocity of 300 μm/ms, based on experimental estimates (Stuart et al., 1997). Furthermore, experimentally measured ratios of NMDA and AMPA conductances were used in order to model their relative contribution to unitary the EPSC (Feldmeyer, 2006; Myme et al., 2003; Rinaldi et al., 2007; Silver et al., 2003; Wang and Gao, 2009).

Reim15

interbouton interval: 1-10 um
syns/bouton: 1 (2 low prob)
bouton density: ~0.2/um
syns/conn: E->E/I ~5 , I->E/I-> ~15

Zado95

attenuation in and out - graphical rep

Huys06

Lond02

Kome09

Jaff99

fig - pyr L5 cell

http://www.neuron.yale.edu/phpbb/viewtopic.php?f=8&t=2404&p=9487&hilit=frequency+response#p9487

http://www.neuron.yale.edu/phpbb/viewtopic.php?f=15&t=313&p=870&hilit=frequency+response#p870

what is

> the significance of this Z transfer for neural communication?

Very simple. Transfer impedance is the best predictor of the effect of synaptic location on synaptic efficacy. This is a consequence of these two facts:

Peak depolarization at the synaptic trigger zone is the

primary determinant of whether or not an epsp will trigger a spike. This is easily shown by computational modeling.

Most synapses act like current sources, not voltage sources.

Also easily shown by computational modeling.

Therefore, despite everything you might find in textbooks, hear in the classroom, or read in most journal articles, voltage attenuation is not a useful predictor of the effect of synaptic location on synaptic efficacy. The best predictor is transfer impedance, which tells you how strongly a current, injected at one point in the cell, will affect membrane potential throughout the cell. See Jaffe, D.B. and Carnevale, N.T. Passive normalization of synaptic integration influenced by dendritic architecture. Journal of Neurophysiology 82:3268-3285, 1999.

So if you want to understand how the distribution of synaptic inputs over the surface of a cell will affect the spiking output of that cell, study the spatial variation of transfer impedance from a reference point located at the cell’s spike trigger zone (since transfer impedance between any two points is independent of the direction of signal propagation, the transfer impedance from any point to the soma is the same as from the soma to that point).

https://github.com/davidcsterratt/dcs-hoc

https://www.neuron.yale.edu/phpBB/viewtopic.php?f=8&t=2264

PointProcessDistributor.hoc http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1002545

16may19 Meeting with Padraig

mod

most channels already exist in NeuroML2
converting NeuroML2 -> mod is possible

github

waffle - organize issues
pull request to release

neuroml -> netpyne

if make changes then post comment or pull request in show case: https://github.com/OpenSourceBrain/NetPyNEShowcase/tree/master/NetPyNE

16jun01 Setting up Travis

example code

Framework for running automated tests on OSB using Travis-CI, see https://github.com/OpenSourceBrain/osb-model-validation

sudo: false

addons: apt: packages:

python-numpy
python-scipy
python-matplotlib
python-sympy

language: python python: 2.7

virtualenv: system_site_packages: true

env:

OMV_ENGINE=PyNEURON
OMV_ENGINE=jNeuroML
OMV_ENGINE=jNeuroML_NEURON
OMV_ENGINE=jNeuroML_validate

install:

pip install git+https://github.com/OpenSourceBrain/osb-model-validation

omv install Brian

omv install NEURON
cd NEURON/test
/home/travis/neuron/nrn/x86_64/bin/nrnivmodl ../mod.files
ls -alt
cd ../..

script:

omv all -V

useful repos

https://github.com/openworm/org.geppetto.simulator.external/blob/master/.travis.yml https://github.com/NeuralEnsemble/PyNN/blob/master/ci/install_neuron.sh https://github.com/OpenSourceBrain/SmithEtAl2013-L23DendriticSpikes/blob/master/.travis.yml

16jun15 New format for figures

Possible formats

list of dict

figs=[]

simConfig[‘analysis’].append({‘func’: ‘plotRaster’, ‘include’: [‘all’, 120, ‘E1’, (‘L2’, 56), (‘L5’, [4,5,6]), (‘L6’, range(5,20))], # all, gid, pop, pop rel index ‘maxSpikes’:3e8, ‘overlaySpikeHist’: True, ‘syncLines’: True, ‘orderBy’:’ynorm’|’y’|’pop’|’cellType’, ‘figId’: 1, # figs with same figId will be converted to subplots ‘saveData’: ‘data2.pkl’, ‘saveName’: ‘fig1.png’})

simConfig[‘analysis’].append({‘func’: ‘plotSpikeHist’, ‘include’: [‘all’, 120, ‘E1’, (‘L2’, 56), (‘L5’, [4,5,6]), (‘L6’, range(5,20))], # all, gid, pop, pop rel index ‘binsize’: 5, ‘overlay’: True, # else separate subplots ‘style’: ‘line’|’bar’|’scatter’, ‘yaxis’: ‘rate’|’count’, ‘figId’: 1, # figs with same figId will be converted to subplots ‘saveData’: ‘data1.pkl, ‘saveFig’: ‘fig1.png’})

simConfig[‘analysis’].append({‘func’: ‘plotTraces’, ‘include’: [‘all’, 120, ‘E1’, (‘L2’, 56), (‘L5’, [4,5,6]), (‘L6’, range(5,20))], # all, gid, pop, pop rel index (automatically added to recordCells) ‘overlay’: True, # else separate subplots ‘oneFigPer’: ‘cell’|’trace’, ‘saveData’: ‘data1.pkl, ‘saveFig’: ‘fig1.png’})

just dict

simConfig[‘analysis’][‘plotRaster’]={ ‘include’: [‘all’, 120, ‘E1’, (‘L2’, 56), (‘L5’, [4,5,6]), (‘L6’, range(5,20))], # all, gid, pop, pop rel index ‘maxSpikes’:3e8, ‘overlaySpikeHist’: True, ‘syncLines’: True, ‘orderBy’:’ynorm’|’y’|’pop’|’cellType’, ‘figId’: 1, # figs with same figId will be converted to subplots ‘saveData’: ‘data2.pkl’, ‘saveName’: ‘fig1.png’})

simConfig[‘analysis’][‘plotConnMatrix’]= True OR {‘values’: ‘weight’| ‘numConns’| ’ |’probability’| ‘convergence’, ‘divergence’, ‘groupBy’: ‘cell’|’pop’, ‘saveData’: ‘data1.pkl, ‘saveFig’: ‘fig1.png’})

Equivalent to calling function

can use same args for calling function or for simConfig (use **kwargs)

plotRaster y-axis – possible code to show actual y values instead of indices

”’maxY = max(y2ind.values()) minY = min(y2ind.values())

base=10 upperY = base * ceil(float(maxY)/base) lowerY = base * floor(float(minY)/base) yAddUpper = int(upperY - maxY) yAddLower = int (minY-lowerY) for i in range(yAddUpper): y2ind.update({yAddUpper: len(y2ind)+1}) for i in range(yAddLower): y2ind.update({yAddLower: min(y2ind.values())-1}) ystep = base #int(len(y2ind)/base) #ystep = base * round(float(ystep)/base) yticks(y2ind.values()[::ystep], y2ind.keys()[::ystep])”’

16Jun16 Parallelizing creation of cells and connections

Email from mike hines

Parallel NEURON idioms: Information exchange during setup Random numbers Debugging Michael Hines CodeJam 2014 Information exchange during setup Results must be independent of Number of processors Distribution of cells Information exchange during setup Results must be independent of Number of processors Distribution of cells A process is often interested in all the objects with a particular property. Information exchange during setup Results must be independent of Number of processors Distribution of cells A process is often interested in all the objects with a particular property. But it generally does not know where the objects are. And the process that owns the object does not know who is interested in it. Information exchange during setup Results must be independent of Number of processors Distribution of cells A process is often interested in all the objects with a particular property. But it generally does not know where the objects are. And the process that owns the object does not know who is interested in it. There is not enough memory in any one process to hold a map of which ranks hold which objects. Example: MPI_ISend/Recv spike exchange Cells do not know which ranks are interested in its spikes. Example: MPI_ISend/Recv spike exchange Cells do not know which ranks are interested in its spikes. Example: Source/Target connectivity Reciprocal synapse connection description. (mitral_gid, mdend_index, xm, granule_gid, gdend_index, xg, …) Construct a mitral => all the tuples with that mitral_gid. Granules don t know enough for construction of the tuples. Construct a granule => gather all the tuples with that granule_gid. Basic exchange: dest = ParallelContext.py_alltoall(src) src and dest are a list of nhost pickleable objects. src[j] on the ith rank will be copied to dest[i] on the jth rank. Likely identical to mpi4py.MPI comm.alltoall(src, dest). 1 2 3 1 2 3 Basic exchange: dest = ParallelContext.py_alltoall(src) src and dest are a list of nhost pickleable objects. src[j] on the ith rank will be copied to dest[i] on the jth rank. Likely identical to mpi4py.MPI comm.alltoall(src, dest). Essentially a wrapper for: MPI_Alltoallv(s, scnt, sdispl, MPI_CHAR, r, rcnt, rdispl, MPI_CHAR, comm); along with a preliminary MPI_all2all(scnt, 1, MPI_INT, rcnt, 1, MPI_INT, comm); in order to calculate rcnt and rdispl. But: No one knows who holds what. No room for anyone to have a global map. But: No one knows who holds what. No room for anyone to have a global map. Solution: A rendezvous rank function: rank = rendezvous(property) usually rank = gid % nhost But: No one knows who holds what. No room for anyone to have a global map. Solution: A rendezvous rank function: rank = rendezvous(property) usually rank = gid % nhost

Everyone sends the keys they own to the rendezvous rank.
Everyone sends the keys they want to the rendezvous rank.
The rendezvous rank sends back to the owners,

which ranks want which keys.

The owners send the objects to the ranks that want them.

Usually simplification is possible: If the objects are small.

Everyone sends the keys and objects

they own to the rendezvous rank.

Everyone sends the keys they want to the rendezvous rank.
The rendezvous rank sends the objects to the ranks that

want them. Usually simplification is possible: If rendezvous(property) is known to be the source rank for all the keys (a priori or by verifying with an all_reduce).

Everyone sends the keys they want to the owner ranks.
The owners send the objects to the ranks that want them.

Usually simplification is possible: If rendezvous(property) is known to be the destination rank for all the keys (a priori or by verifying with an all_reduce).

The owners send the objects to the ranks that want them.

What about RANDOM? Results must be independent of Number of processors Distribution of cells What about RANDOM? Results must be independent of Number of processors Distribution of cells Associate a random stream with a cell. Reproducible Independent Restartable What about RANDOM? Results must be independent of Number of processors Distribution of cells Associate a random stream with a cell. Reproducible Independent Restartable Use cryptographic transformation of several integers. run number stream number (cell gid) stream pick index Use cryptographic transformation of several integers. run number stream number (cell gid) stream pick index Had been using MCellRan4 but only two integers to define x(n1, n2) Use cryptographic transformation of several integers. run number stream number (cell gid) stream pick index Had been using MCellRan4 but only two integers to define x(n1, n2) Thanks! to Eilif Muller for suggesting: Parallel Random Numbers: As Easy as 1, 2, 3 Salmon et al. SC11 (2011) D. E. Shaw Research, New York, NY 10036, USA We introduce several counter based PRNGs: some based on cryptographic standards (AES, Threefish) and some completely new (Philox). All our PRNGs pass rigorous statistical tests (including TestU01 s BigCrush) and produce at least 2^64 unique parallel streams of random numbers, each with period 2^128 or more. http://www.deshawresearch.com/resources_random123.html Use cryptographic transformation of several integers. run number stream number (cell gid) stream pick index Random123: Eight integers define x(n1, …, n8) But we use 5 (three for the stream number). Philox variant Use cryptographic transformation of several integers. run number stream number (cell gid) stream pick index Random123: Eight integers define x(n1, …, n8) But we use 5 (three for the stream number). Philox variant Good performance 10 million picks ACG 0.329s MLCG 0.681s MCellRan4 0.150s numpy.random.rand(n) 0.233s Random123 0.201s (Mersenne Twister) from neuron import h r = h.Random() r.Random123(1,2,3) r.negexp(1) from neuron import h r = h.Random() r.Random123(1,2,3) r.negexp(1) 1e+06 1e+06 800000 995000 600000 990000 400000 985000 0 0.005 0.01 0.015 200000 n=100000000 0 dx = .01 0 1 2 y = h.Vector(n).setrand(r) y = y.histogram(0,5,dx).rotate( 1,0) x = y.c().indgen(dx/2,dx) g = h.Graph() y.line(g, x) 3 4 5 nrnran123.h (abridged) all generator instances share the global index extern void nrnran123_set_globalindex(uint32_t gix); extern nrnran123_State* nrnran123_newstream(uint32_t id1, uint32_t id2); extern uint32_t nrnran123_ipick(nrnran123_State*); uniform 0 to 2^32 1 extern double nrnran123_dblpick(nrnran123_State*); uniform open interval (0,1) minimum value is 2.3283064e 10 max value is 1 min extern double nrnran123_negexp(nrnran123_State*); mean 1.0 min value is 2.3283064e 10 max is 22.18071 extern double nrnran123_negexp(nrnran123_State*); mean 1.0 min value is 2.3283064e 10 max is 22.18071 log(1/2^32) 22.18071 log(2/2^32) 21.487563 log(10/2^32) 19.878125 log(11/2^32) 19.782815 exp( 5)*2^32 28939262 log(28939262/2^32) 5.0000000001 log(28939263/2^32) 4.99999996 extern double nrnran123_negexp(nrnran123_State*); mean 1.0 min value is 2.3283064e 10 max is 22.18071 stateless (though the global index is still used) extern nrnran123_array4x32 nrnran123_iran(uint32_t seq, uint32_t id1, uint32_t id2); Debugging Results must be independent of Number of processors Distribution of cells Debugging Results must be independent of Number of processors Distribution of cells

GID and time of first spike difference.

Debugging Results must be independent of Number of processors Distribution of cells

GID and time of first spike difference.
All spikes delivered to synapses of that Cell?

Debugging Results must be independent of Number of processors Distribution of cells

GID and time of first spike difference.
All spikes delivered to synapses of that Cell?
When and what is the first state difference?

h.load_file( prcellstate.hoc ) if pc.gid_exists(gid): h.prcellgid(gid)

16jul11 Modify network instance

sim.net.modifyCells

sim.net.modifyConns

sim.net.modifyStims

16jul15 Use dot notation

new classes Dict and ODict

~25% slower

old dicts

Creating simulation of 23 cell populations for 0.1 s on 1 hosts… Number of cells on node 0: 4768 Done; cell creation time = 3.07 s. Making connections… Number of connections on node 0: 9536 Done; cell connection time = 1.28 s.

Running… 0 Done; run time = 125.21 s; real-time ratio: 0.00.

Gathering spikes… Done; gather time = 1.91 s.

Analyzing… Run time: 125.21 s Simulated time: 0.1 s; 4768 cells; 1 workers Spikes: 2120 (4.45 Hz) Connections: 9536 (2.00 per cell) 162.0 89.4 820 IT_L23 2.20985431331 Hz 89.0 89.4 113 PV_L23 8.80996218645 Hz 97.0 89.4 113 SOM_L23 9.60186889984 Hz 69.0 89.4 431 IT_L4 1.79074728663 Hz 72.0 89.4 475 IT_L5A 1.69551395267 Hz 111.0 89.4 540 IT_L5B 2.29927914492 Hz 751.0 89.4 540 PT_L5B 15.5563841246 Hz 169.0 89.4 248 PV_L5 7.6225012629 Hz 235.0 89.4 248 SOM_L5 10.5993360756 Hz 83.0 89.4 496 IT_L6 1.8717976474 Hz 95.0 89.4 496 CT_L6 2.14241899401 Hz 81.0 89.4 124 PV_L6 7.30677635852 Hz 106.0 89.4 124 SOM_L6 9.56195424695 Hz Plotting recorded cell traces … Done; plotting time = 0.61 s

Total time = 132.20 s

new dot Dicts

Creating simulation of 23 cell populations for 0.1 s on 1 hosts… Number of cells on node 0: 4768 Done; cell creation time = 4.08 s. Making connections… Number of connections on node 0: 9536 Done; cell connection time = 1.55 s.

Running… 0 Done; run time = 123.59 s; real-time ratio: 0.00.

Gathering spikes… Done; gather time = 3.34 s.

Analyzing… Run time: 123.59 s Simulated time: 0.1 s; 4768 cells; 1 workers Spikes: 2120 (4.45 Hz) Connections: 9536 (2.00 per cell) 162.0 89.4 820 IT_L23 2.20985431331 Hz 89.0 89.4 113 PV_L23 8.80996218645 Hz 97.0 89.4 113 SOM_L23 9.60186889984 Hz 69.0 89.4 431 IT_L4 1.79074728663 Hz 72.0 89.4 475 IT_L5A 1.69551395267 Hz 111.0 89.4 540 IT_L5B 2.29927914492 Hz 751.0 89.4 540 PT_L5B 15.5563841246 Hz 169.0 89.4 248 PV_L5 7.6225012629 Hz 235.0 89.4 248 SOM_L5 10.5993360756 Hz 83.0 89.4 496 IT_L6 1.8717976474 Hz 95.0 89.4 496 CT_L6 2.14241899401 Hz 81.0 89.4 124 PV_L6 7.30677635852 Hz 106.0 89.4 124 SOM_L6 9.56195424695 Hz Plotting recorded cell traces … Done; plotting time = 0.93 s

Total time = 133.78 s

16jul19 Allen Brain example

emails

— Hi Salvador,

Thank you very much for talking to us today. That was helpful.

As a next step for us, I wonder if I could ask you to give a quick tutorial to a larger group of folks here at the Allen Institute. We have a meeting with several more people planned for Wednesday, July 20, at 3 pm Pacific time. Would you be able to talk to us then?

I am thinking that perhaps you could speak for 20 minutes or so, and then do 10 minutes of Q&A.

It would be most helpful if for this tutorial you could walk us through a specific example of NetPyNE workflow applied to a question that’s of interest to us. It doesn’t have to be a fully finished example – could be pseudo-code in part. We would just really benefit from seeing how you are doing these operations.

Here’s the example I have in mind.

Let’s say you have 5 or so neuron types, each represented by a particular cell model and a morphology. Let’s build a model consisting of 10,000 neurons that are replicas of these 5 models.
Distribute the 10,000 cells in space.
Connect the neurons using cell-type specific connectivity rules (let’s say, probabilistic rules based on distances between somata).
Establish feedforward inputs into the 10,000 cells representing spike trains incoming from other parts of the brain. Use arbitrary spike trains (e.g., experimental recordings), if possible.
Save the constructed system to a file. Read the file into NetPyNE and run the simulation.
Visualize the output.

Do you think it’s feasible? Sorry for a short notice, and please do let me know if this is too much to ask :).

Many thanks,

Anton.

— Hi Anton,

Yes, I think its feasible.

Were you interested in any particular cell models / number of compartments? any ones available in hoc or python (eg. from ModelDB) can be directly imported. Of course 10k cells with very detailed morphologies might take a while to simulate. In our M1 sim we combine Izhikevich point neurons, 5-compartment, and full >170 compartment cells. Let me know if you’d be interested in seeing an example of this.

Any requirements in terms of subcellular connectivity? Want single soma to soma synapse, or multiple synapses per cell-to-cell connection? Currently can provide specific section/location for each synapse, or can choose to distribute uniformly across a set of sections or sectionList (also about to release feature to distribute synapses over dendritic tree based on more complex patterns, eg. 1D or 2D density maps).

I’ll let you know if I have any issues, but otherwise talk to you Wednesday 3pm (pacific time).

Thanks again for your interest. Salva

— Sounds great, Salvador. Thank you very much.

We typically use models with a few hundred compartments. They all can be found here: http://celltypes.brain-map.org/ They are also available on ModelDB.

Let’s not worry about actually running a simulation with 10,000 cells. We are just curious to see the workflow and get an idea about the scale of the problem. If you can estimate how long it will take to run the model-building part, how long it would take to simulate it, and how large are the files in which you would save the model details, that would be great, but don’t worry if it turns out that estimating those numbers is not straightforward.

For the connectivity, we are using multiple synapses per connection (let’s assume, five synapses per connection on average). Synapses should be distributed over dendritic tree, with rules depending on target and source cell types. For example, something like “for this connection type, place synapses on the soma and dendrites within 100 um of path distance from the soma”. Again, please don’t worry about actually instantiating all this. We would just like to see an example of your code that does that.

One additional thought we have is that we could ask you for the code from such a tutorial after we talk on Wednesday. That way we could start playing around with NetPyNE in a meaningful way. Hope that’s OK.

Thanks, and looking forward to chatting soon! Best,

Anton.

results

300x1200x300

create_save.py:

Creating network of 9 cell populations on 1 hosts… Number of cells on node 0: 6477 Done; cell creation time = 11.53 s. Making connections… Number of connections on node 0: 4160680 Done; cell connection time = 617.51 s. Adding stims… Number of stims on node 0: 32385 Done; cell stims creation time = 3.49 s.

Gathering data… Done; gather time = 193.44 s.

Analyzing… Cells: 6477 Connections: 4193065 (647.38 per cell) Saving output as Allen.json … Finished saving!

size: 970Mb

load_run.py:

Loading file Allen.json … Done; file loading time = 56.78 s Loading simConfig… simConfig not found in file Allen.json Loading netParams… Loading net… Created 6477 cells Added NEURON objects to 6477 cells Done; re-instantiate net time = 75.16 s Loading simData… simData not found in file Allen.json Recording V_soma from cell 0 Recording V_dend6 from cell 0 Recording V_soma from cell 6045 Recording V_dend6 from cell 6045

380x120x380

create_save.py:

Creating network of 9 cell populations on 1 hosts… Number of cells on node 0: 10392 Done; cell creation time = 17.84 s. Making connections… Number of connections on node 0: 9647815 Done; cell connection time = 1625.43 s. Adding stims… Number of stims on node 0: 51960 Done; cell stims creation time = 6.23 s.

Gathering data… Done; gather time = 563.89 s.

Analyzing… Cells: 10392 Connections: 9699775 (933.39 per cell) Saving output as Allen.json … Finished saving!

size: 2.0Gb (crashed)

380x120x380 (less conns)

create_save.py:

Creating network of 9 cell populations on 1 hosts… Number of cells on node 0: 10392 Done; cell creation time = 18.68 s. Making connections… Number of connections on node 0: 6329300 Done; cell connection time = 1388.36 s. Adding stims… Number of stims on node 0: 51960 Done; cell stims creation time = 6.06 s.

Gathering data… Done; gather time = 353.89 s.

Analyzing… Cells: 10392 Connections: 6381260 (614.06 per cell) Saving output as Allen.json … Finished saving!

size: 1.4Gb (crashed)

load_run.py:

Loading file Allen_10k.json … Done; file loading time = 99.93 s Loading simConfig… simConfig not found in file Allen_10k.json Loading netParams… Loading net… Created 10392 cells Created 6381260 connections Created 51960 stims Added NEURON objects to 10392 cells Done; re-instantiate net time = 116.31 s Loading simData… simData not found in file Allen_10k.json

Running… Done; run time = 32.55 s; real-time ratio: 0.00.

Gathering data… Done; gather time = 751.81 s.

Analyzing… Cells: 10392 Connections: 6381260 (614.06 per cell) Spikes: 0 (0.00 Hz) Simulated time: 0.0 s; 1 workers Run time: 32.55 s Done; saving time = 0.11 s. Plotting raster… No spikes available to plot raster Plotting recorded cell traces … Plotting 2D representation of network cell locations and connections…

380x120x380 (less conns, with subcell)

16jul20 Matplotlib errors

http://stackoverflow.com/questions/4130355/python-matplotlib-framework-under-macosx
fixed by adding ‘backend: Agg’ to ~/.matplotlib/matplotlibrc
actually just deactivated

16aug20 Fixing memory leak when running batch

links

http://chase-seibert.github.io/blog/2013/08/03/diagnosing-memory-leaks-python.html
http://stackoverflow.com/questions/11891755/find-all-references-to-an-object-in-python/11891904#11891904
http://stackoverflow.com/questions/10446839/does-dictionarys-clear-method-delete-all-the-item-related-objects-from-memory
http://stackoverflow.com/questions/7101404/how-can-i-release-memory-after-creating-matplotlib-figures
https://pymotw.com/2/gc/

testing different cleaning methods

added after iteration:

print ‘Memory usage: %s’ % resource.getrusage(resource.RUSAGE_SELF).ru_maxrss

memory increases in master node after iter

master: 89 mb -> 362 -> 569 -> 800 -> 993 -> 1190 slaves: 90 -> 139 …

trying gc.collect() - same
tried: replaceItemObj([cell.__dict__ for cell in sim.net.cells], ‘h’, None)

lowered mem, but crashed – need to gid_clear before

with gid_clear before, increasing but less

98 -> 180 -> 230

clear recorded stuff

89 -> 180 -> 230 -> 280

trying:

import objgraph objgraph.show_most_common_types()

function 11471 Dict 10633 list 8925 dict 6759 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094

function 11471 Dict 10658 list 8929 dict 6761 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094

function 11471 Dict 10597 list 8917 dict 6755 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094 Dict 63291 list 19046

function 11471 dict 6786 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094 Setting stimulation (NetCon) weight to 0.02 Memory usage: 181530624 Memory usage: 140632064 Memory usage: 140087296 Memory usage: 140316672

function 11471 list 11433 Dict 10658 dict 9270 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094

function 11471 list 11421 Dict 10597 dict 9264 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094

function 11471 list 11429 Dict 10633 dict 9268 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094 Dict 113384 list 31602

function 11471 dict 9303 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094 Setting stimulation (NetCon) weight to 0.03 Memory usage: 234979328 Memory usage: 140632064 Memory usage: 140087296 Memory usage: 140316672

clear allSimData and allCells

function 11471 Dict 10597 list 8917 dict 6755 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094

function 11471 Dict 10658 list 8929 dict 6761 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094

function 11471 Dict 10633 list 8925 dict 6759 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094 Dict 63291 list 19011

function 11471 dict 6785 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094 Setting stimulation (NetCon) weight to 0.02 Memory usage: 183537664 Memory usage: 140525568 Memory usage: 140050432 Memory usage: 140353536

function 11471 list 11421 Dict 10597 dict 9264 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094

function 11471 list 11429 Dict 10633 dict 9268 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094

function 11471 list 11433 Dict 10658 dict 9270 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094 Dict 115886 list 31567

function 11471 dict 9302 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094 Setting stimulation (NetCon) weight to 0.03 Memory usage: 233951232 Memory usage: 140353536 Memory usage: 140525568 Memory usage: 140050432

remove allCells and simDAta objects themselves

Memory usage: 183980032 Dict 52598 list 16498 function 11471 dict 5529 tuple 2616 wrapper_descriptor 1880 weakref 1748 builtin_function_or_method 1288 method_descriptor 1194 type 1094

Memory usage: 232386560 Dict 105193 list 29054 function 11471 dict 8046 tuple 2481 wrapper_descriptor 1895 weakref 1752 builtin_function_or_method 1288 method_descriptor 1196 type 1094

objects in sim: ‘allSimData’, ‘simData’, ‘net’, ‘cfg’
if no gather, then only increases 2mb / iteration
only happens when multiple cores! related to gather
leakage when pc.py_alltoall – need to clear vars inside func
small leakage when saving - think reduced by clearing vars
huge leakge with plotSpikeHist

16nov23 Adding support for gap junctions

useful links

http://www.neuron.yale.edu/neuron/static/new_doc/modelspec/programmatic/network/parcon.html?highlight=gap%20junction https://senselab.med.yale.edu/ModelDB/ShowModel.cshtml?model=97917 https://senselab.med.yale.edu/ModelDB/ShowModel.cshtml?model=97917

example from padraig

if (isCellOnNode(“SampleCellGroup”, 1)) { a_SampleCellGroup[1].Soma { elecsyn_NetConn_SampleCellGroup_SampleCellGroup_ElectSyn_A[0] = new ElectSyn(0.5) } elecsyn_NetConn_SampleCellGroup_SampleCellGroup_ElectSyn_A[0].weight = 1.0 pnm.pc.target_var(&elecsyn_NetConn_SampleCellGroup_SampleCellGroup_ElectSyn_A[0].vgap, 100000000) pnm.pc.source_var(&a_SampleCellGroup[1].Soma.v(0.5), 200000000) } if (isCellOnNode(“SampleCellGroup”, 0)) { a_SampleCellGroup[0].Soma { elecsyn_NetConn_SampleCellGroup_SampleCellGroup_ElectSyn_B[0] = new ElectSyn(0.5) } elecsyn_NetConn_SampleCellGroup_SampleCellGroup_ElectSyn_B[0].weight = 1.0 pnm.pc.target_var(&elecsyn_NetConn_SampleCellGroup_SampleCellGroup_ElectSyn_B[0].vgap, 200000000) pnm.pc.source_var(&a_SampleCellGroup[0].Soma.v(0.5), 100000000) }

issues

bidirectional
requires creating synMechs both in presyn and postsyn
requires creating conns from pre->post and post->pre
need to do at the network.py level so can call createConn func for both cells
requires specifying sec and loc of source cell as well as of target cell – add preSec and preLoc
create a temporal list with the presyn cells and ids, and create the synMech, target_var and source_var with the appropriate ids after connections have been created.
check that it plays well with subcellular distributions
check that no duplicates are created

16oct26 Elliptic cylinder and ellipsoid shapes

if sim.net.params.shape == ‘cylinder’:

rho = randLocs[:,0] # use x rand value as the radius rho in the interval [0, 1) phi = 2 * pi * randLocs[:,2] # use z rand value as the angle phi in the interval [0, 2*pi) x = (1 + sqrt(rho) * cos(phi))/2.0 z = (1 + sqrt(rho) * sin(phi))/2.0 randLocs[:,0] = x randLocs[:,2] = z

elif sim.net.params.shape == ‘ellipsoid’:

rho = np.power(randLocs[:,0], 1.0/3.0) # use x rand value as the radius rho in the interval [0, 1); cuberoot phi = 2 * pi * randLocs[:,1] # use y rand value as the angle phi in the interval [0, 2*pi) costheta = (2 * randLocs[:,2]) - 1 # use z rand value as cos(theta) in the interval [-1, 1); ensures uniform dist theta = arccos(costheta) # obtain theta from cos(theta) x = (1 + rho * cos(phi) * sin(theta))/2.0 y = (1 + rho * sin(phi) * sin(theta))/2.0 z = (1 + rho * cos(theta))/2.0 randLocs[:,0] = x randLocs[:,1] = y randLocs[:,2] = z

16oct27 Netstims and artif cells as cell populations

change tutorials
if artificial cells (spike generators with no section), ‘cellModel’ refers to point process mech name (can be overriden if cellRule with that name)
check if cellModel=rule, otherwise check if pointprocess, otherwise error
use try except to set params
create oneToOne connectivity – connect in order
backward compatibility issues:

– netstim pop now requires numCells (default to numCells: 1 ??) – connection should be oneToOne (default is fullConn)

cant check if rules exist for particular cellModel, since could be applied based on other features (eg. yfrac); so have to make condition of point process make prioritary: 1st check if cellModel==point process -> PointCell; otherwise CompartCell

17Apr05 Error checking

popParams

cellModel: required, arbitrary value or pointprocess;

– if arbitrary then create compartCell() and use cellParams – if pointtprocess then create pointCell() and don’t need cellParams

cellType: not required, but recommended; arbitrary value; used to associate population with cellParam rules
numCells, density or gridSpacing - at least 1 required

– numCells: fixed number of cells (integer) – density: num cells per mm3 (float) – gridSpacing: fixed spacing between cells (um)

xRange, yRange, zRange: optional, absolute allowed range to place cells; check within network size
xnormRange, ynormRange, znormRange: optional, normalized allowed range to place cells; check within 0,1
for NetStims:

rate - Firing rate in Hz (note this is the inverse of the NetStim interval property). noise - Fraction of noise in NetStim (0 = deterministic; 1 = completely random). number - Max number of spikes generated (default = 1e12) seed - Seed for randomizer (optional; defaults to value set in simConfig.seeds[‘stim’])

check netParams network dimensions:

– netParams.sizeX, sizeY, sizeZ - net dimensions in um; check not 0 or negative – shape - check either ‘cuboid’, ‘cylinder’ or ‘ellipsoid’

cellParams

conds: required and needs to be a dict eg. (‘conds’: {‘pop’: [‘PYR2’,’PYR3’], ‘x’:[100,200]})

– keys: attributes/tags of cell (cell.tags) — declared in popParams: eg. cellType, cellModel, any arbitrary — popLabel — x, y, z — xnorm, ynorm, znorm

– values: — popLabel, cellModel, arbitrary attributes: warning if doensn’t exist in popParams; single string or list of strings — x, y, z: should be within 0,sizeX etc – check format is single value of 2-value list: — xnorm, ynorm, znorm: should be within range 0,1

secs

– geom: required; either pt3d or L+diam – topol: required if len(secs)>1; has to be dict; with 3 keys: — parentSec: string which should be in keys of secs — parentX: 0-1 — childX: 0-1

– mechs: not required, check mech and properties exist using utils.mechVarList()

– ions: not required, check mech and properties exist using utils.mechVarList()

– pointps: not required; – required key ‘mod’ wiht pointp label; check exists using utils.mechVarlist() – ‘loc’ also required; has to be between 0 and 1 – ‘vref’ and ‘synList’ optional

spikeGenLoc: not rquired; between 0 and 1

stimTargetParams

source: needs to be one of the keys/labels in netParams.stimSourceParams

connParams

preConds - same as cellParams conds
postConds - same as cellParams conds
sec: optional; uses ‘soma’ by default, 1st section available; check if section exists in at least 1 cell type – cellParams (otherwise need to ‘calculate’ postConds and might not make sense in error checking stage)
loc: optional; defaults to 0.5; has to be [0,1];

– can be a list if synMechs is a list, or synsPerConn > 1 – size of list = len(synMechs) or synsPerConn – each item in list in range [0,1] – if both len(synMechs) and synsPerConn>1, loc can be a 2D list (e.g. for 2 synMechs and synsPerConn = 3: [[0.2, 0.3, 0.5], [0.5, 0.6, 0.7]])

weight: optional; any arbitrary positive value; single value, list or 2d list (see loc above)
delay: optional; any arbitrary positive value; single value, list or 2d list (see loc above)
synMech: optional; defaults to synMechParams[0]; can be single string or list; items have to be a key in netParams.synMechParams
synsPerConn: optional, defaults to 1; has to be >=1
probability: optional; [0,1]
convergence/divergence: optional; positive integer
connFunc: deprecated
string-based functions - see valid expressions in documentation

todo

flags

cfg.checkErrors - true or false (affects simFuncs initialize()) cfg.checkErrorsVerbose

True - prints evrything, passed and failed
False - only failed, including warnings (nothing for passed)

add extra tests for numeric values eg. numCells

error messages not capitalized

XNormRange -> xnormRange etc

stop if an error happens

sys.exit

17Apr07 GUI

gantt chart

start before sept
have minor demo (pics) by Oct so can include in grant – maybe move standalone to later
import cells should be together with netPArams UI
should all fit within avail budget – if not, give us tasks that we can do

gantt chart

infrastructure

improv design - required
migrate to npm - started, dependencies managed, needs to happen and easier to maintain, required
update to latest jupyter - doesn’t need to happen
improve interaction - need to understand
sync mechs - almost 0
decouple components - postpone until see if required, maybe refactor in future
refactor + bug fixing - cell builder issue

standalone application - nice to have

transition - will become smaller

NetPyNE UI

netPAramsUI + simConfigUI - for chickens

Network support

map geppetoo - required
performance to load bigger models - 10k
visualization - simplify 3 different levels
advance - remove
ability to simulate large nets
visualize results - focus on netpyne/matplotlib, and use Neuron-UI voltage trace

Support for parallel simulation

run from commandline
maybe don’t need - just use to sketch out

embed python UI into Geppetto

nice to have, but not essential

bug fixing

new proposal

hourly rate: $75/hr - gross days = estimate at 4.5hr/day managing - 15%

simulate large nets + visualize nets - related to mpi, reduce
transition from high -> instantite - room for simplification
project - 55
design - linear

17Apr13 convert to python3

doc

https://docs.python.org/2/library/2to3.html

command

2to3 –output-dir=python3-version/mycode -W -n python2-version/mycode

output of 2to3

Salvador-Duras-MacBook-Pro% 2to3 –output-dir=netpyne_py3 -W -n netpyne WARNING: –write-unchanged-files/-W implies -w. lib2to3.main: Output in ‘netpyne_py3’ will mirror the input directory ‘netpyne’ layout. RefactoringTool: Skipping optional fixer: buffer RefactoringTool: Skipping optional fixer: idioms RefactoringTool: Skipping optional fixer: set_literal RefactoringTool: Skipping optional fixer: ws_comma RefactoringTool: No changes to netpyne/__init__.py RefactoringTool: Writing converted netpyne/__init__.py to netpyne_py3/__init__.py. RefactoringTool: Refactored netpyne/analysis.py — netpyne/analysis.py (original) +++ netpyne/analysis.py (refactored) @@ -17,7 +17,7 @@ from numbers import Number import math

-import sim +from . import sim

import warnings warnings.filterwarnings(“ignore”) @@ -40,7 +40,7 @@ sim.timing(‘start’, ‘plotTime’)

for funcName, kwargs in sim.cfg.analysis.iteritems():
for funcName, kwargs in sim.cfg.analysis.items(): if kwargs == True: kwargs = {} elif kwargs == False: continue func = getattr(sim.analysis, funcName) # get pointer to function

@@ -50,12 +50,12 @@ if sim.cfg.timing:

sim.timing(‘stop’, ‘plotTime’)

print(’ Done; plotting time = %0.2f s’ % sim.timingData[‘plotTime’])
print((’ Done; plotting time = %0.2f s’ % sim.timingData[‘plotTime’]))
sim.timing(‘stop’, ‘totalTime’)
sumTime = sum([t for k,t in sim.timingData.iteritems() if k not in [‘totalTime’]])
sumTime = sum([t for k,t in sim.timingData.items() if k not in [‘totalTime’]]) if sim.timingData[‘totalTime’] <= 1.2*sumTime: # Print total time (only if makes sense)
print(‘\nTotal time = %0.2f s’ % sim.timingData[‘totalTime’])
print((‘\nTotal time = %0.2f s’ % sim.timingData[‘totalTime’]))

###################################################################################################################################################### @@ -72,22 +72,22 @@ ## Save figure data ###################################################################################################################################################### def _saveFigData(figData, fileName=None, type=”):

if not fileName or not isinstance(fileName, basestring):
if not fileName or not isinstance(fileName, str): fileName = sim.cfg.filename+’_’~~type~~‘.pkl’
if fileName.endswith(‘.pkl’): # save to pickle import pickle
print(‘Saving figure data as %s … ’ % (fileName))
print((‘Saving figure data as %s … ’ % (fileName))) with open(fileName, ‘wb’) as fileObj: pickle.dump(figData, fileObj)
elif fileName.endswith(‘.json’): # save to json import json
print(‘Saving figure data as %s … ’ % (fileName))
print((‘Saving figure data as %s … ’ % (fileName))) with open(fileName, ‘w’) as fileObj: json.dump(figData, fileObj) else:
print ‘File extension to save figure data not recognized’
print(‘File extension to save figure data not recognized’)

import numpy @@ -129,10 +129,10 @@ “””

if x.ndim != 1:

raise ValueError, “smooth only accepts 1 dimension arrays.”
raise ValueError(“smooth only accepts 1 dimension arrays.”)
if x.size < window_len:
raise ValueError, “Input vector needs to be bigger than window size.”
raise ValueError(“Input vector needs to be bigger than window size.”)

if window_len<3: @@ -140,7 +140,7 @@

if not window in [‘flat’, ‘hanning’, ‘hamming’, ‘bartlett’, ‘blackman’]:

raise ValueError, “Window is on of ‘flat’, ‘hanning’, ‘hamming’, ‘bartlett’, ‘blackman’”
raise ValueError(“Window is on of ‘flat’, ‘hanning’, ‘hamming’, ‘bartlett’, ‘blackman’”)

s=numpy.r_[x[window_len-1:0:-1],x,x[-1:-window_len:-1]] @@ -173,7 +173,7 @@ ###################################################################################################################################################### def getCellsInclude(include): allCells = sim.net.allCells

allNetStimLabels = sim.net.params.stimSourceParams.keys()
allNetStimLabels = list(sim.net.params.stimSourceParams.keys()) cellGids = [] cells = [] netStimLabels = []

@@ -194,7 +194,7 @@ elif isinstance(condition, int): # cell gid cellGids.append(condition)

elif isinstance(condition, basestring): # entire pop
elif isinstance(condition, str): # entire pop if condition in allNetStimLabels: netStimLabels.append(condition) else:

@@ -260,7 +260,7 @@ if len(cellGids) > 0: gidColors = {cell[‘gid’]: popColors[cell[‘tags’][‘pop’]] for cell in cells} # dict with color for each gid try:

spkgids,spkts = zip(*[(spkgid,spkt) for spkgid,spkt in zip(sim.allSimData[‘spkid’],sim.allSimData[‘spkt’]) if spkgid in cellGids])
spkgids,spkts = list(zip(*[(spkgid,spkt) for spkgid,spkt in zip(sim.allSimData[‘spkid’],sim.allSimData[‘spkt’]) if spkgid in cellGids])) except: spkgids, spkts = [], [] spkgidColors = [gidColors[spkgid] for spkgid in spkgids]

@@ -294,8 +294,8 @@ numCellSpks = len(spkts) numNetStims = 0 for netStimLabel in netStimLabels:

netStimSpks = [spk for cell,stims in sim.allSimData[‘stims’].iteritems() \
for stimLabel,stimSpks in stims.iteritems() for spk in stimSpks if stimLabel == netStimLabel]
netStimSpks = [spk for cell,stims in sim.allSimData[‘stims’].items() \
for stimLabel,stimSpks in stims.items() for spk in stimSpks if stimLabel == netStimLabel] if len(netStimSpks) > 0: lastInd = max(spkinds) if len(spkinds)>0 else 0 spktsNew = netStimSpks

@@ -314,7 +314,7 @@ ylabelText = ylabelText + ‘NetStims’

if numCellSpks+numNetStims == 0:

print ‘No spikes available to plot raster’
print(‘No spikes available to plot raster’) return None

@@ -323,14 +323,14 @@ elif timeRange is None: timeRange = [0,sim.cfg.duration] else:

spkinds,spkts,spkgidColors = zip(*[(spkind,spkt,spkgidColor) for spkind,spkt,spkgidColor in zip(spkinds,spkts,spkgidColors)
if timeRange[0] <= spkt <= timeRange[1]])
spkinds,spkts,spkgidColors = list(zip(*[(spkind,spkt,spkgidColor) for spkind,spkt,spkgidColor in zip(spkinds,spkts,spkgidColors)
if timeRange[0] <= spkt <= timeRange[1]]))
if (len(spkts)>maxSpikes):
print(’ Showing only the first %i out of %i spikes’ % (maxSpikes, len(spkts))) # Limit num of spikes
print((’ Showing only the first %i out of %i spikes’ % (maxSpikes, len(spkts)))) # Limit num of spikes if numNetStims: # sort first if have netStims
spkts, spkinds, spkgidColors = zip(*sorted(zip(spkts, spkinds, spkgidColors)))
spkts, spkinds, spkgidColors = list(zip(*sorted(zip(spkts, spkinds, spkgidColors)))) spkts = spkts[:maxSpikes] spkinds = spkinds[:maxSpikes] spkgidColors = spkgidColors[:maxSpikes]

@@ -446,7 +446,7 @@

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’_’+’raster.png’

@@ -499,7 +499,7 @@ if yaxis == ‘rate’: yaxisLabel = ‘Avg cell firing rate (Hz)’ elif yaxis == ‘count’: yaxisLabel = ‘Spike count’ else:

print ‘Invalid yaxis value %s’, (yaxis)
print(‘Invalid yaxis value %s’, (yaxis)) return

@@ -520,7 +520,7 @@

if len(cellGids) > 0: try:

spkinds,spkts = zip(*[(spkgid,spkt) for spkgid,spkt in zip(sim.allSimData[‘spkid’],sim.allSimData[‘spkt’]) if spkgid in cellGids])
spkinds,spkts = list(zip(*[(spkgid,spkt) for spkgid,spkt in zip(sim.allSimData[‘spkid’],sim.allSimData[‘spkt’]) if spkgid in cellGids])) except: spkinds,spkts = [],[] else:

@@ -531,8 +531,8 @@ numNetStims = 0 if ‘stims’ in sim.allSimData: for netStimLabel in netStimLabels:

netStimSpks = [spk for cell,stims in sim.allSimData[‘stims’].iteritems() \
for stimLabel,stimSpks in stims.iteritems() for spk in stimSpks if stimLabel == netStimLabel]
netStimSpks = [spk for cell,stims in sim.allSimData[‘stims’].items() \
for stimLabel,stimSpks in stims.items() for spk in stimSpks if stimLabel == netStimLabel] if len(netStimSpks) > 0: lastInd = max(spkinds) if len(spkinds)>0 else 0 spktsNew = netStimSpks

@@ -590,7 +590,7 @@

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’_’+’spikeHist.png’

@@ -660,7 +660,7 @@

if len(cellGids) > 0: try:

spkinds,spkts = zip(*[(spkgid,spkt) for spkgid,spkt in zip(sim.allSimData[‘spkid’],sim.allSimData[‘spkt’]) if spkgid in cellGids])
spkinds,spkts = list(zip(*[(spkgid,spkt) for spkgid,spkt in zip(sim.allSimData[‘spkid’],sim.allSimData[‘spkt’]) if spkgid in cellGids])) except: spkinds,spkts = [],[] else:

@@ -672,8 +672,8 @@ numNetStims = 0 if ‘stims’ in sim.allSimData: for netStimLabel in netStimLabels:

netStimSpks = [spk for cell,stims in sim.allSimData[‘stims’].iteritems() \
for stimLabel,stimSpks in stims.iteritems() for spk in stimSpks if stimLabel == netStimLabel]
netStimSpks = [spk for cell,stims in sim.allSimData[‘stims’].items() \
for stimLabel,stimSpks in stims.items() for spk in stimSpks if stimLabel == netStimLabel] if len(netStimSpks) > 0: lastInd = max(spkinds) if len(spkinds)>0 else 0 spktsNew = netStimSpks

@@ -736,7 +736,7 @@

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’_’+’spikePSD.png’

@@ -786,7 +786,7 @@ sim.setupRecording() sim.simulate()

tracesList = sim.cfg.recordTraces.keys()
tracesList = list(sim.cfg.recordTraces.keys()) tracesList.sort() cells, cellGids, _ = getCellsInclude(include) gidPops = {cell[‘gid’]: cell[‘tags’][‘pop’] for cell in cells}

@@ -833,7 +833,7 @@ if ‘cell_’+str(gid) in sim.allSimData[trace]: fullTrace = sim.allSimData[trace][‘cell_’+str(gid)] if isinstance(fullTrace, dict):

data = [fullTrace[key][int(timeRange[0]/recordStep):int(timeRange[1]/recordStep)] for key in fullTrace.keys()]
data = [fullTrace[key][int(timeRange[0]/recordStep):int(timeRange[1]/recordStep)] for key in list(fullTrace.keys())] lenData = len(data[0]) data = np.transpose(array(data)) else:

@@ -862,7 +862,7 @@

elif oneFigPer == ‘popTrace’: allPopGids = invertDictMapping(gidPops)

for popLabel, popGids in allPopGids.iteritems():
for popLabel, popGids in allPopGids.items(): plotFigPerTrace(popGids)
try:

@@ -879,12 +879,12 @@

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’_’+’traces.png’ if len(figs) > 1:
for figLabel, figObj in figs.iteritems():
for figLabel, figObj in figs.items(): plt.figure(figObj.number) plt.savefig(filename[:-4]+figLabel+filename[-4:]) else:

@@ -898,7 +898,7 @@ def invertDictMapping(d): “”” Invert mapping of dictionary (i.e. map values to list of keys) “”” inv_map = {}

for k, v in d.iteritems():
for k, v in d.items(): inv_map[v] = inv_map.get(v, []) inv_map[v].append(k) return inv_map

@@ -951,7 +951,7 @@

if cvar == ‘weightNorm’: for cellPost in cellsPost:

cellSecs = cellPost.secs.values() if includeAxon else [s for s in cellPost.secs.values() if ‘axon’ not in s[‘hSec’].hname()]
cellSecs = list(cellPost.secs.values()) if includeAxon else [s for s in list(cellPost.secs.values()) if ‘axon’ not in s[‘hSec’].hname()] for sec in cellSecs: if ‘weightNorm’ in sec: secs.append(sec[‘hSec’])

@@ -963,8 +963,8 @@

elif cvar == ‘numSyns’: for cellPost in cellsPost:

cellSecs = cellPost.secs if includeAxon else {k:s for k,s in cellPost.secs.iteritems() if ‘axon’ not in s[‘hSec’].hname()}
for secLabel,sec in cellSecs.iteritems():
cellSecs = cellPost.secs if includeAxon else {k:s for k,s in cellPost.secs.items() if ‘axon’ not in s[‘hSec’].hname()}
for secLabel,sec in cellSecs.items(): nseg=sec[‘hSec’].nseg nsyns = [0] * nseg secs.append(sec[‘hSec’])

@@ -974,7 +974,7 @@

cvals = np.array(cvals)

if not secs: secs = [s[‘hSec’] for cellPost in cellsPost for s in cellPost.secs.values()]
if not secs: secs = [s[‘hSec’] for cellPost in cellsPost for s in list(cellPost.secs.values())]

@@ -998,7 +998,7 @@ if showSyns: synColor=’red’ for cellPost in cellsPost:

for sec in cellPost.secs.values():
for sec in list(cellPost.secs.values()): for synMech in sec[‘synMechs’]: morph.mark_locations(h, sec[‘hSec’], synMech[‘loc’], markspec=synStyle, color=synColor, markersize=synSiz)

@@ -1007,7 +1007,7 @@

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’_shape.png’

@@ -1024,7 +1024,7 @@ ivprops = {‘colorSecs’: 1, ‘colorSyns’:2 ,’style’: ‘.’, ‘siz’:10}

for cell in [c for c in sim.net.cells if c.tags[‘pop’] in includePost]:

for sec in cell.secs.values():
for sec in list(cell.secs.values()): if ‘axon’ in sec[‘hSec’].hname() and not includeAxon: continue sec[‘hSec’].push() secList.append()

@@ -1045,7 +1045,7 @@ fig.show(0) # show real diam

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’_’+’shape.ps’

@@ -1136,7 +1136,7 @@ else: cellsPost, cellGidsPost, netStimPopsPost = getCellsInclude(includePost)

if isinstance(synMech, basestring): synMech = [synMech] # make sure synMech is a list
if isinstance(synMech, str): synMech = [synMech] # make sure synMech is a list
if groupBy == ‘cell’:

@@ -1144,7 +1144,7 @@ connMatrix = np.zeros((len(cellGidsPre), len(cellGidsPost))) countMatrix = np.zeros((len(cellGidsPre), len(cellGidsPost))) else:

print ‘Conn matrix with groupBy=”cell” only supports features= “weight”, “delay” or “numConns”’
print(‘Conn matrix with groupBy=”cell” only supports features= “weight”, “delay” or “numConns”’) return fig cellIndsPre = {cell[‘gid’]: ind for ind,cell in enumerate(cellsPre)} cellIndsPost = {cell[‘gid’]: ind for ind,cell in enumerate(cellsPost)}

@@ -1273,7 +1273,7 @@

elif groupBy in sim.net.allCells[0][‘tags’] and isinstance(sim.net.allCells[0][‘tags’][groupBy], Number): if not isinstance(groupByInterval, Number):

print ‘groupByInterval not specified’
print(‘groupByInterval not specified’) return

@@ -1294,7 +1294,7 @@

if len(groupsPre) < 2 or len(groupsPost) < 2:

print ‘groupBy %s with groupByInterval %s results in <2 groups’%(str(groupBy), str(groupByInterval))
print(‘groupBy %s with groupByInterval %s results in <2 groups’%(str(groupBy), str(groupByInterval))) return groupIndsPre = {group: ind for ind,group in enumerate(groupsPre)} groupIndsPost = {group: ind for ind,group in enumerate(groupsPost)}

@@ -1361,7 +1361,7 @@

else:

print ‘groupBy (%s) is not valid’%(str(groupBy))
print(‘groupBy (%s) is not valid’%(str(groupBy))) return
if groupBy != ‘cell’:

@@ -1418,8 +1418,8 @@ plt.plot(array([ipop,ipop])-0.5,array([0,len(popsPost)])-0.5,’-‘,c=(0.7,0.7,0.7))

h.set_xticks(range(len(popsPost)))
h.set_yticks(range(len(popsPre)))
h.set_xticks(list(range(len(popsPost))))
h.set_yticks(list(range(len(popsPre)))) h.set_xticklabels(popsPost) h.set_yticklabels(popsPre) h.xaxis.set_ticks_position(‘top’)

@@ -1463,10 +1463,10 @@ plt.tight_layout()

elif groupBy == ‘cell’:

print ‘Error: plotConn graphType=”bar” with groupBy=”cell” not implemented’
print(‘Error: plotConn graphType=”bar” with groupBy=”cell” not implemented’)
elif graphType == ‘pie’:
print ‘Error: plotConn graphType=”pie” not yet implemented’
print(‘Error: plotConn graphType=”pie” not yet implemented’)

#save figure data @@ -1478,7 +1478,7 @@

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’’+’conn’~~feature~~‘.png’

@@ -1536,7 +1536,7 @@ if showConns: for postCell in cells: for con in postCell[‘conns’]: # plot connections between cells

if not isinstance(con[‘preGid’], basestring) and con[‘preGid’] in cellGids:
if not isinstance(con[‘preGid’], str) and con[‘preGid’] in cellGids: posXpre,posYpre = next(((cell[‘tags’][‘x’],cell[‘tags’][ycoord]) for cell in cells if cell[‘gid’]==con[‘preGid’]), None) posXpost,posYpost = postCell[‘tags’][‘x’], postCell[‘tags’][ycoord] color=’red’

@@ -1566,7 +1566,7 @@

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’_’+’2Dnet.png’

@@ -1603,13 +1603,13 @@

if ‘nte’ not in dir(h): try:

print ’ Warning: support/nte.mod not compiled; attempting to compile from %s via “nrnivmodl support”’%(root)
print(’ Warning: support/nte.mod not compiled; attempting to compile from %s via “nrnivmodl support”’%(root)) os.system(‘cd ’ + root + ‘; nrnivmodl support’) from neuron import load_mechanisms load_mechanisms(root)
print ’ Compilation of support folder mod files successful’
print(’ Compilation of support folder mod files successful’) except:
print ’ Error compiling support folder mod files’
print(’ Error compiling support folder mod files’) return
h.load_file(root+’/support/nte.hoc’) # nTE code (also requires support/net.mod)

@@ -1632,7 +1632,7 @@ numNetStims = 0 for netStimPop in netStimPops: if ‘stims’ in sim.allSimData:

cellStims = [cellStim for cell,cellStim in sim.allSimData[‘stims’].iteritems() if netStimPop in cellStim]
cellStims = [cellStim for cell,cellStim in sim.allSimData[‘stims’].items() if netStimPop in cellStim] if len(cellStims) > 0: spktsNew = [spkt for cellStim in cellStims for spkt in cellStim[netStimPop] ] spkts.extend(spktsNew)

@@ -1658,7 +1658,7 @@ numNetStims = 0 for netStimPop in netStimPops: if ‘stims’ in sim.allSimData:

cellStims = [cellStim for cell,cellStim in sim.allSimData[‘stims’].iteritems() if netStimPop in cellStim]
cellStims = [cellStim for cell,cellStim in sim.allSimData[‘stims’].items() if netStimPop in cellStim] if len(cellStims) > 0: spktsNew = [spkt for cellStim in cellStims for spkt in cellStim[netStimPop] ] spkts.extend(spktsNew)

@@ -1737,7 +1737,7 @@ numNetStims = 0 for netStimPop in netStimPops: if ‘stims’ in sim.allSimData:

cellStims = [cellStim for cell,cellStim in sim.allSimData[‘stims’].iteritems() if netStimPop in cellStim]
cellStims = [cellStim for cell,cellStim in sim.allSimData[‘stims’].items() if netStimPop in cellStim] if len(cellStims) > 0: spktsNew = [spkt for cellStim in cellStims for spkt in cellStim[netStimPop] ] spkts.extend(spktsNew)

@@ -1763,7 +1763,7 @@ numNetStims = 0 for netStimPop in netStimPops: if ‘stims’ in sim.allSimData:

cellStims = [cellStim for cell,cellStim in sim.allSimData[‘stims’].iteritems() if netStimPop in cellStim]
cellStims = [cellStim for cell,cellStim in sim.allSimData[‘stims’].items() if netStimPop in cellStim] if len(cellStims) > 0: spktsNew = [spkt for cellStim in cellStims for spkt in cellStim[netStimPop] ] spkts.extend(spktsNew)

@@ -1808,7 +1808,7 @@

if saveFig:

if isinstance(saveFig, basestring):
if isinstance(saveFig, str): filename = saveFig else: filename = sim.cfg.filename+’_’+’2Dnet.png’

@@ -1841,7 +1841,7 @@

wcs = [x[-1][-1] for x in sim.allweightchanges] # absolute final weight wcs = [x[-1][-1]-x[0][-1] for x in sim.allweightchanges] # absolute weight change

pre,post,recep = zip(*[(x[0],x[1],x[2]) for x in sim.allstdpconndata])
pre,post,recep = list(zip(*[(x[0],x[1],x[2]) for x in sim.allstdpconndata])) ncells = int(max(max(pre),max(post))+1) wcmat = np.zeros([ncells, ncells])

RefactoringTool: Writing converted netpyne/analysis.py to netpyne_py3/analysis.py. RefactoringTool: Refactored netpyne/batch.py — netpyne/batch.py (original) +++ netpyne/batch.py (refactored) @@ -8,7 +8,7 @@

import datetime -from itertools import izip, product +from itertools import product from popen2 import popen2 from time import sleep import imp @@ -22,11 +22,11 @@ def runJob(script, cfgSavePath, netParamsSavePath): from subprocess import Popen, PIPE

print ‘\nJob in rank id: ‘,pc.id()
print(‘\nJob in rank id: ‘,pc.id()) command = ‘nrniv %s simConfig=%s netParams=%s’ % (script, cfgSavePath, netParamsSavePath)
print command+’\n’
print(command+’\n’) proc = Popen(command.split(’ ‘), stdout=PIPE, stderr=PIPE)
print proc.stdout.read()
print(proc.stdout.read())

class Batch(object): @@ -40,7 +40,7 @@ self.runCfg = {} self.params = [] if params:

for k,v in params.iteritems():
for k,v in params.items(): self.params.append({‘label’: k, ‘values’: v})
def save(self, filename):

@@ -54,14 +54,14 @@ os.mkdir(folder) except OSError: if not os.path.exists(folder):

print ’ Could not create’, folder
print(’ Could not create’, folder)
dataSave = {‘batch’: self.__dict__} if ext == ‘json’: import json #from json import encoder #encoder.FLOAT_REPR = lambda o: format(o, ‘.12g’)
print(‘Saving batch to %s … ’ % (filename))
print((‘Saving batch to %s … ’ % (filename))) with open(filename, ‘w’) as fileObj: json.dump(dataSave, fileObj, indent=4, sort_keys=True)

@@ -74,7 +74,7 @@ os.mkdir(self.saveFolder) except OSError: if not os.path.exists(self.saveFolder):

print ’ Could not create’, self.saveFolder
print(’ Could not create’, self.saveFolder)
targetFile = self.saveFolder+’/’~~self.batchLabel~~’_batch.json’

@@ -102,14 +102,14 @@ elif p[‘group’] == True: groupedParams = True

labelList, valuesList = zip(*[(p[‘label’], p[‘values’]) for p in self.params if p[‘group’] == False])
labelList, valuesList = list(zip(*[(p[‘label’], p[‘values’]) for p in self.params if p[‘group’] == False])) valueCombinations = list(product(*(valuesList)))
indexCombinations = list(product(*[range(len(x)) for x in valuesList]))
indexCombinations = list(product(*[list(range(len(x))) for x in valuesList]))
if groupedParams:
labelListGroup, valuesListGroup = zip(*[(p[‘label’], p[‘values’]) for p in self.params if p[‘group’] == True])
valueCombGroups = izip(*(valuesListGroup))
indexCombGroups = izip(*[range(len(x)) for x in valuesListGroup])
labelListGroup, valuesListGroup = list(zip(*[(p[‘label’], p[‘values’]) for p in self.params if p[‘group’] == True]))
valueCombGroups = zip(*(valuesListGroup))
indexCombGroups = zip(*[list(range(len(x))) for x in valuesListGroup]) labelList = labelListGroup+labelList else: valueCombGroups = [(0,)] # this is a hack – improve!

@@ -133,7 +133,7 @@ iComb = iCombNG pComb = pCombNG

print iComb, pComb
print(iComb, pComb)
for i, paramVal in enumerate(pComb): paramLabel = labelList[i]

@@ -147,7 +147,7 @@ container[paramLabel[-1]] = paramVal else: setattr(self.cfg, paramLabel, paramVal) # set simConfig params

print str(paramLabel)+’ = ‘+str(paramVal)
print(str(paramLabel)+’ = ‘+str(paramVal))
simLabel = self.batchLabel+”.join([”.join(‘_’+str(i)) for i in iComb])

@@ -155,11 +155,11 @@

if self.runCfg.get(‘skip’, False) and glob.glob(jobName+’.json’):

print ‘Skipping job %s since output file already exists…’ % (jobName)
print(‘Skipping job %s since output file already exists…’ % (jobName)) elif self.runCfg.get(‘skipCfg’, False) and glob.glob(jobName+’_cfg.json’):
print ‘Skipping job %s since cfg file already exists…’ % (jobName)
print(‘Skipping job %s since cfg file already exists…’ % (jobName)) elif self.runCfg.get(‘skipCustom’, None) and glob.glob(jobName+self.runCfg[‘skipCustom’]):
print ‘Skipping job %s since %s file already exists…’ % (jobName, self.runCfg[‘skipCustom’])
print(‘Skipping job %s since %s file already exists…’ % (jobName, self.runCfg[‘skipCustom’])) else:
self.cfg.simLabel = simLabel

@@ -199,7 +199,7 @@

input.write(jobString)

print jobString+’\n’
print(jobString+’\n’) input.close()

@@ -239,8 +239,8 @@

print ‘Submitting job ‘,jobName
print jobString+’\n’
print(‘Submitting job ‘,jobName)
print(jobString+’\n’)
batchfile = ‘%s.sbatch’%(jobName)

@@ -255,7 +255,7 @@

elif self.runCfg.get(‘type’,None) == ‘mpi’: jobName = self.saveFolder+’/’+simLabel

print ‘Submitting job ‘,jobName
print(‘Submitting job ‘,jobName)
pc.submit(runJob, self.runCfg.get(‘script’, ‘init.py’), cfgSavePath, netParamsSavePath)

RefactoringTool: Writing converted netpyne/batch.py to netpyne_py3/batch.py. RefactoringTool: Refactored netpyne/cell.py — netpyne/cell.py (original) +++ netpyne/cell.py (refactored) @@ -10,10 +10,10 @@ from copy import deepcopy from time import sleep from neuron import h # Import NEURON -from specs import Dict +from .specs import Dict import numpy as np from random import seed, uniform -import sim +from . import sim

############################################################################### @@ -89,22 +89,22 @@ self.stims.append(Dict(params.copy())) # add new stim to Cell object stimContainer = self.stims[-1]

if sim.cfg.verbose: print(’ Created %s NetStim for cell gid=%d’% (params[‘source’], self.gid))
if sim.cfg.verbose: print((’ Created %s NetStim for cell gid=%d’% (params[‘source’], self.gid)))
if sim.cfg.createNEURONObj: rand = h.Random() stimContainer[‘hRandom’] = rand # add netcon object to dict in conns list
if isinstance(params[‘rate’], basestring):
if isinstance(params[‘rate’], str): if params[‘rate’] == ‘variable’: try: netstim = h.NSLOC() netstim.interval = 0.1**-1*1e3 # inverse of the frequency and then convert from Hz^-1 to ms (set very low) netstim.noise = params[‘noise’] except:
print ‘Error: tried to create variable rate NetStim but NSLOC mechanism not available’
print(‘Error: tried to create variable rate NetStim but NSLOC mechanism not available’) else:
print ‘Error: Unknown stimulation rate type: %s’%(h.params[‘rate’])
print(‘Error: Unknown stimulation rate type: %s’%(h.params[‘rate’])) else: netstim = h.NetStim() netstim.interval = params[‘rate’]**-1*1e3 # inverse of the frequency and then convert from Hz^-1 to ms

@@ -126,12 +126,12 @@

def recordTraces (self):

for key, params in sim.cfg.recordTraces.iteritems():
for key, params in sim.cfg.recordTraces.items():
conditionsMet = 1

if ‘conds’ in params:
for (condKey,condVal) in params[‘conds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in params[‘conds’].items(): # check if all conditions are met
if condKey in [‘postGid’]: compareTo = self.gid

@@ -143,7 +143,7 @@ if compareTo < condVal[0] or compareTo > condVal[1]: conditionsMet = 0 break

elif isinstance(condVal, list) and isinstance(condVal[0], basestring):
elif isinstance(condVal, list) and isinstance(condVal[0], str): if compareTo not in condVal: conditionsMet = 0 break

@@ -172,7 +172,7 @@ else: ptr = [] secLocs = []

for secName,sec in self.secs.iteritems():
for secName,sec in self.secs.items(): synMechs = [synMech for synMech in sec[‘synMechs’] if synMech[‘label’]==params[‘synMech’]] ptr.extend([synMech[‘hSyn’].__getattribute__(’ref’~~params[‘var’]) for synMech in synMechs]) secLocs.extend([secName~~‘_’+str(synMech[‘loc’]) for synMech in synMechs])

@@ -193,11 +193,11 @@ else: sim.simData[key][‘cell_’+str(self.gid)] = h.Vector(sim.cfg.duration/sim.cfg.recordStep+1).resize(0) sim.simData[key][‘cell_’+str(self.gid)].record(ptr, sim.cfg.recordStep)

if sim.cfg.verbose: print ’ Recording ‘, key, ‘from cell ‘, self.gid, ’ with parameters: ‘,str(params)
if sim.cfg.verbose: print(’ Recording ‘, key, ‘from cell ‘, self.gid, ’ with parameters: ‘,str(params)) except:
if sim.cfg.verbose: print ’ Cannot record ‘, key, ‘from cell ‘, self.gid
if sim.cfg.verbose: print(’ Cannot record ‘, key, ‘from cell ‘, self.gid) else:
if sim.cfg.verbose: print ’ Conditions preclude recording ‘, key, ’ from cell ‘, self.gid
if sim.cfg.verbose: print(’ Conditions preclude recording ‘, key, ’ from cell ‘, self.gid) #else:

@@ -222,15 +222,15 @@

def create (self):

for propLabel, prop in sim.net.params.cellParams.iteritems(): # for each set of cell properties
for propLabel, prop in sim.net.params.cellParams.items(): # for each set of cell properties conditionsMet = 1
for (condKey,condVal) in prop[‘conds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in prop[‘conds’].items(): # check if all conditions are met if isinstance(condVal, list): if isinstance(condVal[0], Number): if self.tags.get(condKey) < condVal[0] or self.tags.get(condKey) > condVal[1]: conditionsMet = 0 break
elif isinstance(condVal[0], basestring):
elif isinstance(condVal[0], str): if self.tags.get(condKey) not in condVal: conditionsMet = 0 break

@@ -251,7 +251,7 @@

def modify (self, prop): conditionsMet = 1

for (condKey,condVal) in prop[‘conds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in prop[‘conds’].items(): # check if all conditions are met if condKey==’label’: if condVal not in self.tags[‘label’]: conditionsMet = 0

@@ -261,7 +261,7 @@ if self.tags.get(condKey) < condVal[0] or self.tags.get(condKey) > condVal[1]: conditionsMet = 0 break

elif isinstance(condVal[0], basestring):
elif isinstance(condVal[0], str): if self.tags.get(condKey) not in condVal: conditionsMet = 0 break

@@ -278,7 +278,7 @@

def createPyStruct (self, prop):

for sectName,sectParams in prop[‘secs’].iteritems():
for sectName,sectParams in prop[‘secs’].items():
if sectName not in self.secs: self.secs[sectName] = Dict() # create section dict

@@ -286,22 +286,22 @@

if ‘mechs’ in sectParams:

for mechName,mechParams in sectParams[‘mechs’].iteritems():
for mechName,mechParams in sectParams[‘mechs’].items(): if ‘mechs’ not in sec: sec[‘mechs’] = Dict() if mechName not in sec[‘mechs’]: sec[‘mechs’][mechName] = Dict()
for mechParamName,mechParamValue in mechParams.iteritems(): # add params of the mechanism
for mechParamName,mechParamValue in mechParams.items(): # add params of the mechanism sec[‘mechs’][mechName][mechParamName] = mechParamValue
if ‘ions’ in sectParams:
for ionName,ionParams in sectParams[‘ions’].iteritems():
for ionName,ionParams in sectParams[‘ions’].items(): if ‘ions’ not in sec: sec[‘ions’] = Dict() if ionName not in sec[‘ions’]: sec[‘ions’][ionName] = Dict()
for ionParamName,ionParamValue in ionParams.iteritems(): # add params of the ion
for ionParamName,ionParamValue in ionParams.items(): # add params of the ion sec[‘ions’][ionName][ionParamName] = ionParamValue

@@ -313,19 +313,19 @@

if ‘pointps’ in sectParams:

for pointpName,pointpParams in sectParams[‘pointps’].iteritems():
for pointpName,pointpParams in sectParams[‘pointps’].items(): #if self.tags[‘cellModel’] == pointpName: # only required if want to allow setting various cell models in same rule if ‘pointps’ not in sec: sec[‘pointps’] = Dict() if pointpName not in sec[‘pointps’]: sec[‘pointps’][pointpName] = Dict()
for pointpParamName,pointpParamValue in pointpParams.iteritems(): # add params of the mechanism
for pointpParamName,pointpParamValue in pointpParams.items(): # add params of the mechanism sec[‘pointps’][pointpName][pointpParamName] = pointpParamValue

if ‘geom’ in sectParams:

for geomParamName,geomParamValue in sectParams[‘geom’].iteritems():
for geomParamName,geomParamValue in sectParams[‘geom’].items(): if ‘geom’ not in sec: sec[‘geom’] = Dict() if not type(geomParamValue) in [list, dict]: # skip any list or dic params

@@ -342,7 +342,7 @@ if ‘topol’ in sectParams: if ‘topol’ not in sec: sec[‘topol’] = Dict()

for topolParamName,topolParamValue in sectParams[‘topol’].iteritems():
for topolParamName,topolParamValue in sectParams[‘topol’].items(): sec[‘topol’][topolParamName] = topolParamValue

@@ -361,14 +361,14 @@

def initV (self):

for sec in self.secs.values():
for sec in list(self.secs.values()): if ‘vinit’ in sec: sec[‘hSec’].v = sec[‘vinit’]

def createNEURONObj (self, prop):

for sectName,sectParams in prop[‘secs’].iteritems():
for sectName,sectParams in prop[‘secs’].items():
if sectName not in self.secs: self.secs[sectName] = Dict() # create sect dict if doesn’t exist

@@ -378,7 +378,7 @@

if ‘geom’ in sectParams:

for geomParamName,geomParamValue in sectParams[‘geom’].iteritems():
for geomParamName,geomParamValue in sectParams[‘geom’].items(): if not type(geomParamValue) in [list, dict]: # skip any list or dic params setattr(sec[‘hSec’], geomParamName, geomParamValue)

@@ -393,11 +393,11 @@

if ‘mechs’ in sectParams:

for mechName,mechParams in sectParams[‘mechs’].iteritems():
for mechName,mechParams in sectParams[‘mechs’].items(): if mechName not in sec[‘mechs’]: sec[‘mechs’][mechName] = Dict() sec[‘hSec’].insert(mechName)
for mechParamName,mechParamValue in mechParams.iteritems(): # add params of the mechanism
for mechParamName,mechParamValue in mechParams.items(): # add params of the mechanism mechParamValueFinal = mechParamValue for iseg,seg in enumerate(sec[‘hSec’]): # set mech params for each segment if type(mechParamValue) in [list]:

@@ -407,11 +407,11 @@

if ‘ions’ in sectParams:

for ionName,ionParams in sectParams[‘ions’].iteritems():
for ionName,ionParams in sectParams[‘ions’].items(): if ionName not in sec[‘ions’]: sec[‘ions’][ionName] = Dict() sec[‘hSec’].insert(ionName+’_ion’) # insert mechanism
for ionParamName,ionParamValue in ionParams.iteritems(): # add params of the mechanism
for ionParamName,ionParamValue in ionParams.items(): # add params of the mechanism ionParamValueFinal = ionParamValue for iseg,seg in enumerate(sec[‘hSec’]): # set ion params for each segment if type(ionParamValue) in [list]:

@@ -436,20 +436,20 @@

if ‘pointps’ in sectParams:

for pointpName,pointpParams in sectParams[‘pointps’].iteritems():
for pointpName,pointpParams in sectParams[‘pointps’].items(): #if self.tags[‘cellModel’] == pointpParams: # only required if want to allow setting various cell models in same rule if pointpName not in sec[‘pointps’]: sec[‘pointps’][pointpName] = Dict() pointpObj = getattr(h, pointpParams[‘mod’]) loc = pointpParams[‘loc’] if ‘loc’ in pointpParams else 0.5 # set location sec[‘pointps’][pointpName][‘hPointp’] = pointpObj(loc, sec = sec[‘hSec’]) # create h Pointp object (eg. h.Izhi2007b)
for pointpParamName,pointpParamValue in pointpParams.iteritems(): # add params of the point process
for pointpParamName,pointpParamValue in pointpParams.items(): # add params of the point process if pointpParamName not in [‘mod’, ‘loc’, ‘vref’, ‘synList’] and not pointpParamName.startswith(‘_’): setattr(sec[‘pointps’][pointpName][‘hPointp’], pointpParamName, pointpParamValue)

for sectName,sectParams in prop[‘secs’].iteritems(): # iterate sects again for topology (ensures all exist)
for sectName,sectParams in prop[‘secs’].items(): # iterate sects again for topology (ensures all exist) sec = self.secs[sectName] # pointer to section # pointer to child sec if ‘topol’ in sectParams: if sectParams[‘topol’]:

@@ -458,7 +458,7 @@

def addSynMechsNEURONObj(self):

for sectName,sectParams in self.secs.iteritems():
for sectName,sectParams in self.secs.items():
if ‘synMechs’ in sectParams: for synMech in sectParams[‘synMechs’]:

@@ -475,8 +475,8 @@

elif stimParams[‘type’] in [‘IClamp’, ‘VClamp’, ‘SEClamp’, ‘AlphaSynapse’]: stim = getattr(h, stimParams[‘type’])(self.secs[stimParams[‘sec’]][‘hSec’](stimParams[‘loc’]))

stimProps = {k:v for k,v in stimParams.iteritems() if k not in [‘label’, ‘type’, ‘source’, ‘loc’, ‘sec’, ‘h’+stimParams[‘type’]]}
for stimPropName, stimPropValue in stimProps.iteritems(): # set mechanism internal stimParams
stimProps = {k:v for k,v in stimParams.items() if k not in [‘label’, ‘type’, ‘source’, ‘loc’, ‘sec’, ‘h’+stimParams[‘type’]]}
for stimPropName, stimPropValue in stimProps.items(): # set mechanism internal stimParams if isinstance(stimPropValue, list): if stimPropName == ‘amp’: for i,val in enumerate(stimPropValue):

@@ -504,9 +504,9 @@ try: postTarget = synMech[‘hSyn’] except:

print ‘\nError: no synMech available for conn: ‘, conn
print ’ cell tags: ‘,self.tags
print ’ cell synMechs: ‘,self.secs[conn[‘sec’]][‘synMechs’]
print(‘\nError: no synMech available for conn: ‘, conn)
print(’ cell tags: ‘,self.tags)
print(’ cell synMechs: ‘,self.secs[conn[‘sec’]][‘synMechs’]) exit()

@@ -533,15 +533,15 @@ if self.secs: if sim.cfg.createNEURONObj: sim.pc.set_gid2node(self.gid, sim.rank) # this is the key call that assigns cell gid to a particular node

sec = next((secParams for secName,secParams in self.secs.iteritems() if ‘spikeGenLoc’ in secParams), None) # check if any section has been specified as spike generator
sec = next((secParams for secName,secParams in self.secs.items() if ‘spikeGenLoc’ in secParams), None) # check if any section has been specified as spike generator if sec: loc = sec[‘spikeGenLoc’] # get location of spike generator within section else:
sec = self.secs[‘soma’] if ‘soma’ in self.secs else self.secs[self.secs.keys()[0]] # use soma if exists, otherwise 1st section
sec = self.secs[‘soma’] if ‘soma’ in self.secs else self.secs[list(self.secs.keys())[0]] # use soma if exists, otherwise 1st section loc = 0.5 nc = None if ‘pointps’ in sec: # if no syns, check if point processes with ‘vref’ (artificial cell)
for pointpName, pointpParams in sec[‘pointps’].iteritems():
for pointpName, pointpParams in sec[‘pointps’].items(): if ‘vref’ in pointpParams: nc = h.NetCon(sec[‘pointps’][pointpName][‘hPointp’].__getattribute__(’ref‘+pointpParams[‘vref’]), None, sec=sec[‘hSec’]) break

@@ -567,7 +567,7 @@ synMech = next((synMech for synMech in sec[‘synMechs’] if synMech[‘label’]==synLabel and synMech[‘loc’]==loc), None) if not synMech: # if synMech not in section, then create synMech = Dict({‘label’: synLabel, ‘loc’: loc})

for paramName, paramValue in synMechParams.iteritems():
for paramName, paramValue in synMechParams.items(): synMech[paramName] = paramValue sec[‘synMechs’].append(synMech) else:

@@ -583,7 +583,7 @@ if not synMech.get(‘hSyn’): # if synMech doesn’t have NEURON obj, then create synObj = getattr(h, synMechParams[‘mod’]) synMech[‘hSyn’] = synObj(loc, sec=sec[‘hSec’]) # create h Syn object (eg. h.Exp2Syn)

for synParamName,synParamValue in synMechParams.iteritems(): # add params of the synaptic mechanism
for synParamName,synParamValue in synMechParams.items(): # add params of the synaptic mechanism if synParamName not in [‘label’, ‘mod’, ‘selfNetCon’, ‘loc’]: setattr(synMech[‘hSyn’], synParamName, synParamValue) elif synParamName == ‘selfNetcon’: # create self netcon required for some synapses (eg. homeostatic)

@@ -591,7 +591,7 @@ locNetCon = synParamValue.get(‘loc’, 0.5) secNetCon = self.secs.get(secLabelNetCon, None) synMech[‘hNetcon’] = h.NetCon(secNetCon[‘hSec’](locNetCon)._ref_v, synMech[”], sec=secNetCon[‘hSec’])

for paramName,paramValue in synParamValue.iteritems():
for paramName,paramValue in synParamValue.items(): if paramName == ‘weight’: synMech[‘hNetcon’].weight[0] = paramValue elif paramName not in [‘sec’, ‘loc’]:

@@ -605,7 +605,7 @@ conditionsMet = 1 if ‘cellConds’ in params: if conditionsMet:

for (condKey,condVal) in params[‘cellConds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in params[‘cellConds’].items(): # check if all conditions are met
if isinstance(condVal, list): if self.tags.get(condKey) < condVal[0] or self.tags.get(condKey) > condVal[1]:

@@ -616,11 +616,11 @@ break

if conditionsMet:

for secLabel,sec in self.secs.iteritems():
for secLabel,sec in self.secs.items(): for synMech in sec[‘synMechs’]: conditionsMet = 1 if ‘conds’ in params:
for (condKey,condVal) in params[‘conds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in params[‘conds’].items(): # check if all conditions are met
if condKey == ‘sec’: if condVal != secLabel:

@@ -630,7 +630,7 @@ if synMech.get(condKey) < condVal[0] or synMech.get(condKey) > condVal[1]: conditionsMet = 0 break

elif isinstance(condVal, list) and isinstance(condVal[0], basestring):
elif isinstance(condVal, list) and isinstance(condVal[0], str): if synMech.get(condKey) not in condVal: conditionsMet = 0 break

@@ -640,14 +640,14 @@

if conditionsMet: # if all conditions are met, set values for this cell exclude = [‘conds’, ‘cellConds’, ‘label’, ‘mod’, ‘selfNetCon’, ‘loc’]

for synParamName,synParamValue in {k: v for k,v in params.iteritems() if k not in exclude}.iteritems():
for synParamName,synParamValue in {k: v for k,v in params.items() if k not in exclude}.items(): if sim.cfg.createPyStruct: synMech[synParamName] = synParamValue if sim.cfg.createNEURONObj: try: setattr(synMech[‘hSyn’], synParamName, synParamValue) except:
print ‘Error setting %s=%s on synMech’ % (synParamName, str(synParamValue))
print(‘Error setting %s=%s on synMech’ % (synParamName, str(synParamValue)))

@@ -661,7 +661,7 @@

if params[‘preGid’] == self.gid:

if sim.cfg.verbose: print ’ Error: attempted to create self-connection on cell gid=%d, section=%s ‘%(self.gid, params.get(‘sec’))
if sim.cfg.verbose: print(’ Error: attempted to create self-connection on cell gid=%d, section=%s ‘%(self.gid, params.get(‘sec’))) return # if self-connection return

@@ -718,7 +718,7 @@

if sim.cfg.createPyStruct:

connParams = {k:v for k,v in params.iteritems() if k not in [‘synsPerConn’]}
connParams = {k:v for k,v in params.items() if k not in [‘synsPerConn’]} connParams[‘weight’] = weights[i] connParams[‘delay’] = delays[i] if not pointp:

@@ -790,7 +790,7 @@ switchtimes.append(sim.cfg.duration)

switchiter = iter(switchtimes)

switchpairs = zip(switchiter,switchiter)
switchpairs = list(zip(switchiter,switchiter)) for pair in switchpairs:
stimvecs = self._shapeStim(width=float(pulsewidth)/1000.0, isi=float(pulseperiod)/1000.0, weight=params[‘weight’], start=float(pair[0])/1000.0, finish=float(pair[1])/1000.0, stimshape=pulsetype)

@@ -809,8 +809,8 @@ sec = params[‘sec’] if pointp else synMechSecs[i] loc = params[‘loc’] if pointp else synMechLocs[i] preGid = netStimParams[‘source’]+’ NetStim’ if netStimParams else params[‘preGid’]

print(’ Created connection preGid=%s, postGid=%s, sec=%s, loc=%.4g, synMech=%s, weight=%.4g, delay=%.2f, threshold=%s’%
(preGid, self.gid, sec, loc, params[‘synMech’], weights[i], delays[i], threshold))
print((’ Created connection preGid=%s, postGid=%s, sec=%s, loc=%.4g, synMech=%s, weight=%.4g, delay=%.2f, threshold=%s’%
(preGid, self.gid, sec, loc, params[‘synMech’], weights[i], delays[i], threshold)))

def modifyConns (self, params): @@ -818,7 +818,7 @@ conditionsMet = 1

if ‘conds’ in params:

for (condKey,condVal) in params[‘conds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in params[‘conds’].items(): # check if all conditions are met
if condKey in [‘postGid’]: compareTo = self.gid

@@ -830,7 +830,7 @@ if compareTo < condVal[0] or compareTo > condVal[1]: conditionsMet = 0 break

elif isinstance(condVal, list) and isinstance(condVal[0], basestring):
elif isinstance(condVal, list) and isinstance(condVal[0], str): if compareTo not in condVal: conditionsMet = 0 break

@@ -839,13 +839,13 @@ break

if conditionsMet and ‘postConds’ in params:

for (condKey,condVal) in params[‘postConds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in params[‘postConds’].items(): # check if all conditions are met
if isinstance(condVal, list) and isinstance(condVal[0], Number): if self.tags.get(condKey) < condVal[0] or self.tags.get(condKey) > condVal[1]: conditionsMet = 0 break
elif isinstance(condVal, list) and isinstance(condVal[0], basestring):
elif isinstance(condVal, list) and isinstance(condVal[0], str): if self.tags.get(condKey) not in condVal: conditionsMet = 0 break

@@ -854,28 +854,28 @@ break

if conditionsMet and ‘preConds’ in params:

print ‘Warning: modifyConns() does not yet support conditions of presynaptic cells’
print(‘Warning: modifyConns() does not yet support conditions of presynaptic cells’)
if conditionsMet: # if all conditions are met, set values for this cell if sim.cfg.createPyStruct:
for paramName, paramValue in {k: v for k,v in params.iteritems() if k not in [‘conds’,’preConds’,’postConds’]}.iteritems():
for paramName, paramValue in {k: v for k,v in params.items() if k not in [‘conds’,’preConds’,’postConds’]}.items(): conn[paramName] = paramValue if sim.cfg.createNEURONObj:
for paramName, paramValue in {k: v for k,v in params.iteritems() if k not in [‘conds’,’preConds’,’postConds’]}.iteritems():
for paramName, paramValue in {k: v for k,v in params.items() if k not in [‘conds’,’preConds’,’postConds’]}.items(): try: if paramName == ‘weight’: conn[‘hNetcon’].weight[0] = paramValue else: setattr(conn[‘hNetcon’], paramName, paramValue) except:
print ‘Error setting %s=%s on Netcon’ % (paramName, str(paramValue))
print(‘Error setting %s=%s on Netcon’ % (paramName, str(paramValue)))

def modifyStims (self, params): conditionsMet = 1 if ‘cellConds’ in params: if conditionsMet:

for (condKey,condVal) in params[‘cellConds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in params[‘cellConds’].items(): # check if all conditions are met
if isinstance(condVal, list): if self.tags.get(condKey) < condVal[0] or self.tags.get(condKey) > condVal[1]:

@@ -890,13 +890,13 @@ conditionsMet = 1

if ‘conds’ in params:

for (condKey,condVal) in params[‘conds’].iteritems(): # check if all conditions are met
for (condKey,condVal) in params[‘conds’].items(): # check if all conditions are met
if isinstance(condVal, list) and isinstance(condVal[0], Number): if stim.get(condKey) < condVal[0] or stim.get(condKey) > condVal[1]: conditionsMet = 0 break
elif isinstance(condVal, list) and isinstance(condVal[0], basestring):
elif isinstance(condVal, list) and isinstance(condVal[0], str): if stim.get(condKey) not in condVal: conditionsMet = 0 break

@@ -908,13 +908,13 @@ if stim[‘type’] == ‘NetStim’: # for netstims, find associated netcon conn = next((conn for conn in self.conns if conn[‘source’] == stim[‘source’]), None) if sim.cfg.createPyStruct:

for paramName, paramValue in {k: v for k,v in params.iteritems() if k not in [‘conds’,’cellConds’]}.iteritems():
for paramName, paramValue in {k: v for k,v in params.items() if k not in [‘conds’,’cellConds’]}.items(): if stim[‘type’] == ‘NetStim’ and paramName in [‘weight’, ‘delay’, ‘threshold’]: conn[paramName] = paramValue else: stim[paramName] = paramValue if sim.cfg.createNEURONObj:
for paramName, paramValue in {k: v for k,v in params.iteritems() if k not in [‘conds’,’cellConds’]}.iteritems():
for paramName, paramValue in {k: v for k,v in params.items() if k not in [‘conds’,’cellConds’]}.items(): try: if stim[‘type’] == ‘NetStim’: if paramName == ‘weight’:

@@ -924,19 +924,19 @@ else: setattr(stim[‘h’+stim[‘type’]], paramName, paramValue) except:

print ‘Error setting %s=%s on stim’ % (paramName, str(paramValue))
print(‘Error setting %s=%s on stim’ % (paramName, str(paramValue)))

def addStim (self, params):

if not params[‘sec’] or (isinstance(params[‘sec’], basestring) and not params[‘sec’] in self.secs.keys()+self.secLists.keys()):
if sim.cfg.verbose: print ’ Warning: no valid sec specified for stim on cell gid=%d so using soma or 1st available. Existing secs: %s; params: %s’%(self.gid, self.secs.keys(),params)
if not params[‘sec’] or (isinstance(params[‘sec’], str) and not params[‘sec’] in list(self.secs.keys())+list(self.secLists.keys())):
if sim.cfg.verbose: print(’ Warning: no valid sec specified for stim on cell gid=%d so using soma or 1st available. Existing secs: %s; params: %s’%(self.gid, list(self.secs.keys()),params)) if ‘soma’ in self.secs: params[‘sec’] = ‘soma’ # use ‘soma’ if exists elif self.secs:
params[‘sec’] = self.secs.keys()[0] # if no ‘soma’, use first sectiona available
params[‘sec’] = list(self.secs.keys())[0] # if no ‘soma’, use first sectiona available else:
if sim.cfg.verbose: print ’ Error: no Section available on cell gid=%d to add stim’%(self.gid)
if sim.cfg.verbose: print(’ Error: no Section available on cell gid=%d to add stim’%(self.gid)) return
sec = self.secs[params[‘sec’]]

@@ -972,9 +972,9 @@

elif params[‘type’] in [‘IClamp’, ‘VClamp’, ‘SEClamp’, ‘AlphaSynapse’]: stim = getattr(h, params[‘type’])(sec[‘hSec’](params[‘loc’]))

stimParams = {k:v for k,v in params.iteritems() if k not in [‘type’, ‘source’, ‘loc’, ‘sec’, ‘label’]}
stimParams = {k:v for k,v in params.items() if k not in [‘type’, ‘source’, ‘loc’, ‘sec’, ‘label’]} stringParams = ”
for stimParamName, stimParamValue in stimParams.iteritems(): # set mechanism internal params
for stimParamName, stimParamValue in stimParams.items(): # set mechanism internal params if isinstance(stimParamValue, list): if stimParamName == ‘amp’: for i,val in enumerate(stimParamValue):

@@ -989,21 +989,21 @@ self.stims.append(Dict(params)) # add to python structure self.stims[-1][‘h’+params[‘type’]] = stim # add stim object to dict in stims list

if sim.cfg.verbose: print(’ Added %s %s to cell gid=%d, sec=%s, loc=%.4g%s’%
(params[‘source’], params[‘type’], self.gid, params[‘sec’], params[‘loc’], stringParams))
if sim.cfg.verbose: print((’ Added %s %s to cell gid=%d, sec=%s, loc=%.4g%s’%
(params[‘source’], params[‘type’], self.gid, params[‘sec’], params[‘loc’], stringParams)))
else:
if sim.cfg.verbose: print(‘Adding exotic stim (NeuroML 2 based?): %s’% params)
if sim.cfg.verbose: print((‘Adding exotic stim (NeuroML 2 based?): %s’% params)) stim = getattr(h, params[‘type’])(sec[‘hSec’](params[‘loc’]))
stimParams = {k:v for k,v in params.iteritems() if k not in [‘type’, ‘source’, ‘loc’, ‘sec’, ‘label’]}
stimParams = {k:v for k,v in params.items() if k not in [‘type’, ‘source’, ‘loc’, ‘sec’, ‘label’]} stringParams = ”
for stimParamName, stimParamValue in stimParams.iteritems(): # set mechanism internal params
for stimParamName, stimParamValue in stimParams.items(): # set mechanism internal params if isinstance(stimParamValue, list):
print “Can’t set point process paramaters of type vector eg. VClamp.amp[3]”
print(“Can’t set point process paramaters of type vector eg. VClamp.amp[3]”) pass #setattr(stim, stimParamName._ref_[0], stimParamValue[0]) elif ‘originalFormat’ in params:
if sim.cfg.verbose: print(’ originalFormat: %s’%(params[‘originalFormat’]))
if sim.cfg.verbose: print((’ originalFormat: %s’%(params[‘originalFormat’]))) if params[‘originalFormat’]==’NeuroML2_stochastic_input’: rand = h.Random() rand.Random123(params[‘stim_count’], sim.id32(‘%d’%(sim.cfg.seeds[‘stim’])))

@@ -1017,20 ~~1017,20 @@ stringParams = stringParams + ‘, ’ + stimParamName +’=’~~ str(stimParamValue) self.stims.append(params) # add to python structure self.stims[-1][‘h’+params[‘type’]] = stim # add stim object to dict in stims list

if sim.cfg.verbose: print(’ Added %s %s to cell gid=%d, sec=%s, loc=%.4g%s’%
(params[‘source’], params[‘type’], self.gid, params[‘sec’], params[‘loc’], stringParams))
if sim.cfg.verbose: print((’ Added %s %s to cell gid=%d, sec=%s, loc=%.4g%s’%
(params[‘source’], params[‘type’], self.gid, params[‘sec’], params[‘loc’], stringParams)))

def _setConnSections (self, params):

if not params.get(‘sec’) or (isinstance(params.get(‘sec’), basestring) and not params.get(‘sec’) in self.secs.keys()+self.secLists.keys()):
if sim.cfg.verbose: print ’ Warning: no valid sec specified for connection to cell gid=%d so using soma or 1st available’%(self.gid)
if not params.get(‘sec’) or (isinstance(params.get(‘sec’), str) and not params.get(‘sec’) in list(self.secs.keys())+list(self.secLists.keys())):
if sim.cfg.verbose: print(’ Warning: no valid sec specified for connection to cell gid=%d so using soma or 1st available’%(self.gid)) if ‘soma’ in self.secs: params[‘sec’] = ‘soma’ # use ‘soma’ if exists elif self.secs:
params[‘sec’] = self.secs.keys()[0] # if no ‘soma’, use first sectiona available
params[‘sec’] = list(self.secs.keys())[0] # if no ‘soma’, use first sectiona available else:
if sim.cfg.verbose: print ’ Error: no Section available on cell gid=%d to add connection’%(self.gid)
if sim.cfg.verbose: print(’ Error: no Section available on cell gid=%d to add connection’%(self.gid)) sec = -1 # if no Sections available print error and exit return sec

@@ -1042,7 +1042,7 @@ secLabels = [] for i,section in enumerate(secList): if section not in self.secs: # remove sections that dont exist; and corresponding weight and delay

if sim.cfg.verbose: print ’ Error: Section %s not available so removing from list of sections for connection to cell gid=%d’%(self.gid)
if sim.cfg.verbose: print(’ Error: Section %s not available so removing from list of sections for connection to cell gid=%d’%(self.gid)) secList.remove(section) if isinstance(params[‘weight’], list): params[‘weight’].remove(params[‘weight’][i]) if isinstance(params[‘delay’], list): params[‘delay’].remove(params[‘delay’][i])

@@ -1077,7 +1077,7 @@

pointp = None if len(secLabels)==1 and ‘pointps’ in self.secs[secLabels[0]]: # check if point processes with ‘vref’ (artificial cell)

for pointpName, pointpParams in self.secs[secLabels[0]][‘pointps’].iteritems():
for pointpName, pointpParams in self.secs[secLabels[0]][‘pointps’].items(): if ‘vref’ in pointpParams: # if includes vref param means doesn’t use Section v or synaptic mechanisms pointp = pointpName if ‘synList’ in pointpParams:

@@ -1088,7 +1088,7 @@ weightIndex = pointpParams[‘synList’].index(params.get(‘synMech’)) # udpate weight index based pointp synList

if pointp and params[‘synsPerConn’] > 1: # only single synapse per connection rule allowed

if sim.cfg.verbose: print ’ Error: Multiple synapses per connection rule not allowed for cells where V is not in section (cell gid=%d) ‘%(self.gid)
if sim.cfg.verbose: print(’ Error: Multiple synapses per connection rule not allowed for cells where V is not in section (cell gid=%d) ‘%(self.gid)) return -1, weightIndex
return pointp, weightIndex

@@ -1098,11 +1098,11 @@ synsPerConn = params[‘synsPerConn’] if not params.get(‘synMech’): if sim.net.params.synMechParams: # if no synMech specified, but some synMech params defined

synLabel = sim.net.params.synMechParams.keys()[0] # select first synMech from net params and add syn
synLabel = list(sim.net.params.synMechParams.keys())[0] # select first synMech from net params and add syn params[‘synMech’] = synLabel
if sim.cfg.verbose: print ’ Warning: no synaptic mechanisms specified for connection to cell gid=%d so using %s ‘%(self.gid, synLabel)
if sim.cfg.verbose: print(’ Warning: no synaptic mechanisms specified for connection to cell gid=%d so using %s ‘%(self.gid, synLabel)) else: # if no synaptic mechanism specified and no synMech params available
if sim.cfg.verbose: print ’ Error: no synaptic mechanisms available to add conn on cell gid=%d ‘%(self.gid)
if sim.cfg.verbose: print(’ Error: no synaptic mechanisms available to add conn on cell gid=%d ‘%(self.gid)) return -1 # if no Synapse available print error and exit

@@ -1134,7 +1134,7 @@ else: secLengths = [1.0 for s in secList] if sim.cfg.verbose:

print(’ Section lengths not available to distribute synapses in cell %d’%self.gid)
print((’ Section lengths not available to distribute synapses in cell %d’%self.gid))
try: totLength = sum(secLengths)

@@ -1153,7 +1153,7 @@ if plasticity and sim.cfg.createNEURONObj: try: plastMech = getattr(h, plasticity[‘mech’], None)(0, sec=sec[‘hSec’]) # create plasticity mechanism (eg. h.STDP)

for plastParamName,plastParamValue in plasticity[‘params’].iteritems(): # add params of the plasticity mechanism
for plastParamName,plastParamValue in plasticity[‘params’].items(): # add params of the plasticity mechanism setattr(plastMech, plastParamName, plastParamValue) if plasticity[‘mech’] == ‘STDP’: # specific implementation steps required for the STDP mech precon = sim.pc.gid_connect(params[‘preGid’], plastMech); precon.weight[0] = 1 # Send presynaptic spikes to the STDP adjuster

@@ -1166,7 +1166,7 @@ self.conns[-1][‘STDPdata’] = {‘preGid’:params[‘preGid’], ‘postGid’: self.gid, ‘receptor’: weightIndex} # Not used; FYI only; store here just so it’s all in one place if sim.cfg.verbose: print(’ Added STDP plasticity to synaptic mechanism’) except:

print ‘Error: exception when adding plasticity using %s mechanism’ % (plasticity[‘mech’])
print(‘Error: exception when adding plasticity using %s mechanism’ % (plasticity[‘mech’]))

@@ -1199,7 +1199,7 @@ try: self.hPointp = getattr(h, self.tags[‘cellModel’])() except:

print “Error creating point process mechanism %s in cell with gid %d” % (self.tags[‘cellModel’], self.gid)
print(“Error creating point process mechanism %s in cell with gid %d” % (self.tags[‘cellModel’], self.gid)) return

@@ -1208,8 +1208,8 @@ self.params[‘rate’] = uniform(self.params[‘rate’][0], self.params[‘rate’][1])

params = {k: v for k,v in self.params.iteritems()}
for paramName, paramValue in params.iteritems():
params = {k: v for k,v in self.params.items()}
for paramName, paramValue in params.items(): try: if paramName == ‘rate’: self.params[‘interval’] = 1000.0/paramValue

@@ -1289,7 +1289,7 @@ spkTimes = np.cumsum(fixedInterval + negexpInterval) + (start - interval*(1-noise))

else:

print ‘\nError: VecStim num spks per cell > %d’ % (maxReproducibleSpks)
print(‘\nError: VecStim num spks per cell > %d’ % (maxReproducibleSpks)) import sys sys.exit()

@@ -1303,12 +1303,12 @@ elif ‘rate’ in pulse: interval = 1000/pulse[‘rate’] else:

print ‘Error: Vecstim pulse missing “rate” or “interval” parameter’
print(‘Error: Vecstim pulse missing “rate” or “interval” parameter’) return
if any([x not in pulse for x in [‘start’, ‘end’]]):
print ‘Error: Vecstim pulse missing “start” and/or “end” parameter’
print(‘Error: Vecstim pulse missing “start” and/or “end” parameter’) return else: noise = pulse[‘noise’] if ‘noise’ in pulse else 0.0

@@ -1396,7 +1396,7 @@

if params[‘preGid’] == self.gid:

if sim.cfg.verbose: print ’ Error: attempted to create self-connection on cell gid=%d, section=%s ‘%(self.gid, params.get(‘sec’))
if sim.cfg.verbose: print(’ Error: attempted to create self-connection on cell gid=%d, section=%s ‘%(self.gid, params.get(‘sec’))) return # if self-connection return

@@ -1417,7 +1417,7 @@

if sim.cfg.createPyStruct:

connParams = {k:v for k,v in params.iteritems() if k not in [‘synsPerConn’]}
connParams = {k:v for k,v in params.items() if k not in [‘synsPerConn’]} connParams[‘weight’] = weights[i] connParams[‘delay’] = delays[i] if netStimParams:

@@ -1465,7 +1465,7 @@ switchtimes.append(sim.cfg.duration)

switchiter = iter(switchtimes)

switchpairs = zip(switchiter,switchiter)
switchpairs = list(zip(switchiter,switchiter)) for pair in switchpairs:
stimvecs = self._shapeStim(width=float(pulsewidth)/1000.0, isi=float(pulseperiod)/1000.0, weight=params[‘weight’], start=float(pair[0])/1000.0, finish=float(pair[1])/1000.0, stimshape=pulsetype)

@@ -1481,8 +1481,8 @@ sec = params[‘sec’] loc = params[‘loc’] preGid = netStimParams[‘source’]+’ NetStim’ if netStimParams else params[‘preGid’]

print(’ Created connection preGid=%s, postGid=%s, sec=%s, loc=%.4g, synMech=%s, weight=%.4g, delay=%.2f, threshold=%s’%
(preGid, self.gid, sec, loc, params[‘synMech’], weights[i], delays[i],params[‘threshold’]))
print((’ Created connection preGid=%s, postGid=%s, sec=%s, loc=%.4g, synMech=%s, weight=%.4g, delay=%.2f, threshold=%s’%
(preGid, self.gid, sec, loc, params[‘synMech’], weights[i], delays[i],params[‘threshold’])))

def initV (self): @@ -1493,7 +1493,7 @@ try: name(*args,**kwargs) except:

print “Error: Function ‘%s’ not yet implemented for Point Neurons” % name
print(“Error: Function ‘%s’ not yet implemented for Point Neurons” % name) return wrapper

RefactoringTool: Writing converted netpyne/cell.py to netpyne_py3/cell.py. RefactoringTool: Refactored netpyne/network.py — netpyne/network.py (original) +++ netpyne/network.py (refactored) @@ -12,9 +12,9 @@ from time import time, sleep from numbers import Number from copy import copy -from specs import ODict +from .specs import ODict from neuron import h # import NEURON -import sim +from . import sim

class Network (object): @@ -51,7 +51,7 @@

############################################################################### def createPops (self):

for popLabel, popParam in self.params.popParams.iteritems(): # for each set of population paramseters
for popLabel, popParam in self.params.popParams.items(): # for each set of population paramseters self.pops[popLabel] = sim.Pop(popLabel, popParam) # instantiate a new object of class Pop and add to list pop return self.pops

@@ -63,17 +63,17 @@ sim.pc.barrier() sim.timing(‘start’, ‘createTime’) if sim.rank==0:

print(“\nCreating network of %i cell populations on %i hosts…” % (len(self.pops), sim.nhosts))
print((“\nCreating network of %i cell populations on %i hosts…” % (len(self.pops), sim.nhosts)))
for ipop in self.pops.values(): # For each pop instantiate the network cells (objects of class ‘Cell’)
for ipop in list(self.pops.values()): # For each pop instantiate the network cells (objects of class ‘Cell’) newCells = ipop.createCells() # create cells for this pop using Pop method self.cells.extend(newCells) # add to list of cells sim.pc.barrier()
if sim.rank==0 and sim.cfg.verbose: print(‘Instantiated %d cells of population %s’%(len(newCells), ipop.tags[‘pop’]))
print(’ Number of cells on node %i: %i ’ % (sim.rank,len(self.cells)))
if sim.rank==0 and sim.cfg.verbose: print((‘Instantiated %d cells of population %s’%(len(newCells), ipop.tags[‘pop’])))
print((’ Number of cells on node %i: %i ’ % (sim.rank,len(self.cells)))) sim.pc.barrier() sim.timing(‘stop’, ‘createTime’)
if sim.rank == 0 and sim.cfg.timing: print(’ Done; cell creation time = %0.2f s.’ % sim.timingData[‘createTime’])
if sim.rank == 0 and sim.cfg.timing: print((’ Done; cell creation time = %0.2f s.’ % sim.timingData[‘createTime’]))
return self.cells

@@ -94,28 +94,28 @@

sources = self.params.stimSourceParams

for targetLabel, target in self.params.stimTargetParams.iteritems(): # for each target parameter set
for targetLabel, target in self.params.stimTargetParams.items(): # for each target parameter set if ‘sec’ not in target: target[‘sec’] = None # if section not specified, make None (will be assigned to first section in cell) if ‘loc’ not in target: target[‘loc’] = None # if location not specified, make None
source = sources.get(target[‘source’])

postCellsTags = allCellTags
for condKey,condValue in target[‘conds’].iteritems(): # Find subset of cells that match postsyn criteria
for condKey,condValue in target[‘conds’].items(): # Find subset of cells that match postsyn criteria if condKey in [‘x’,’y’,’z’,’xnorm’,’ynorm’,’znorm’]:
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.iteritems() if condValue[0] <= tags.get(condKey, None) < condValue[1]} # dict with post Cell objects} # dict with pre cell tags
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.items() if condValue[0] <= tags.get(condKey, None) < condValue[1]} # dict with post Cell objects} # dict with pre cell tags elif condKey == ‘cellList’: pass elif isinstance(condValue, list):
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.iteritems() if tags.get(condKey, None) in condValue} # dict with post Cell objects
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.items() if tags.get(condKey, None) in condValue} # dict with post Cell objects else:
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.iteritems() if tags.get(condKey, None) == condValue} # dict with post Cell objects
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.items() if tags.get(condKey, None) == condValue} # dict with post Cell objects
if ‘cellList’ in target[‘conds’]: orderedPostGids = sorted(postCellsTags.keys()) gidList = [orderedPostGids[i] for i in target[‘conds’][‘cellList’]]
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.iteritems() if gid in gidList}
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.items() if gid in gidList}
strParams = self._stimStrToFunc(postCellsTags, source, target)

@@ -143,10 +143,10 @@

postCell.addStim(params) # call cell method to add connections

print(’ Number of stims on node %i: %i ’ % (sim.rank, sum([len(cell.stims) for cell in self.cells])))
print((’ Number of stims on node %i: %i ’ % (sim.rank, sum([len(cell.stims) for cell in self.cells])))) sim.pc.barrier() sim.timing(‘stop’, ‘stimsTime’)
if sim.rank == 0 and sim.cfg.timing: print(’ Done; cell stims creation time = %0.2f s.’ % sim.timingData[‘stimsTime’])
if sim.rank == 0 and sim.cfg.timing: print((’ Done; cell stims creation time = %0.2f s.’ % sim.timingData[‘stimsTime’]))
return [cell.stims for cell in self.cells]

@@ -173,7 +173,7 @@ dictVars[‘post_znorm’] = lambda postConds: postConds[‘znorm’]

for k,v in self.params.__dict__.iteritems():
for k,v in self.params.__dict__.items(): if isinstance(v, Number): dictVars[k] = v

@@ -181,7 +181,7 @@ strParams = {} for paramStrFunc in paramsStrFunc: strFunc = params[paramStrFunc] # string containing function

strVars = [var for var in dictVars.keys() if var in strFunc and var+’norm’ not in strFunc] # get list of variables used (eg. post_ynorm or dist_xyz)
strVars = [var for var in list(dictVars.keys()) if var in strFunc and var+’norm’ not in strFunc] # get list of variables used (eg. post_ynorm or dist_xyz) lambdaStr = ‘lambda ’ + ‘,’.join(strVars) +’: ’ + strFunc # convert to lambda function lambdaFunc = eval(lambdaStr)

@@ -190,11 ~~190,11 @@ params[paramStrFunc~~‘FuncVars’] = {strVar: dictVars[strVar] for strVar in strVars}

seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+postCellsTags.keys()[0])))
seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+list(postCellsTags.keys())[0])))
strParams[paramStrFunc+’List’] = {postGid: params[paramStrFunc+’Func’](**{k:v if isinstance(v, Number) else v(postCellTags) for k,v in params[paramStrFunc+’FuncVars’].iteritems()})
for postGid,postCellTags in postCellsTags.iteritems()}
strParams[paramStrFunc+’List’] = {postGid: params[paramStrFunc+’Func’](**{k:v if isinstance(v, Number) else v(postCellTags) for k,v in params[paramStrFunc+’FuncVars’].items()})
for postGid,postCellTags in postCellsTags.items()}
return strParams

@@ -218,7 +218,7 @@ if h.arc3d(ii) >= s: b = ii break

if b == -1: print “an error occurred in pointFromLoc, SOMETHING IS NOT RIGHT”
if b == -1: print(“an error occurred in pointFromLoc, SOMETHING IS NOT RIGHT”)
if h.arc3d(b) == s: # shortcut x, y, z = h.x3d(b), h.y3d(b), h.z3d(b)

@@ -256,7 +256,7 @@ sigma_x2_y2 = gridSigma[i2][j2]

if x1 == x2 or y1 == y2:

print “ERROR in closest grid points: “, secName, x1, x2, y1, y2
print(“ERROR in closest grid points: “, secName, x1, x2, y1, y2) else:
sigma = ((sigma_x1_y1*abs(x2-x)*abs(y2-y) + sigma_x2_y1*abs(x-x1)*abs(y2-y) + sigma_x1_y2*abs(x2-x)*abs(y-y1) + sigma_x2_y2*abs(x-x1)*abs(y-y1))/(abs(x2-x1)*abs(y2-y1)))

@@ -272,7 +272,7 @@ sigma_y2 = gridSigma[j2]

if y1 == y2:

print “ERROR in closest grid points: “, secName, y1, y2
print(“ERROR in closest grid points: “, secName, y1, y2) else:
sigma = ((sigma_y1*abs(y2-y) + sigma_y2*abs(y-y1)) / abs(y2-y1))

@@ -290,7 +290,7 @@ sim.timing(‘start’, ‘subConnectTime’) print(’ Distributing synapses based on subcellular connectivity rules…’)

for subConnParamTemp in self.params.subConnParams.values(): # for each conn rule or parameter set
for subConnParamTemp in list(self.params.subConnParams.values()): # for each conn rule or parameter set subConnParam = subConnParamTemp.copy()

@@ -302,7 +302,7 @@ if postCellGid in self.lid2gid: postCell = self.cells[self.gid2lid[postCellGid]] allConns = [conn for conn in postCell.conns if conn[‘preGid’] in preCellsTags]

if ‘NetStim’ in [x[‘cellModel’] for x in preCellsTags.values()]: # temporary fix to include netstim conns
if ‘NetStim’ in [x[‘cellModel’] for x in list(preCellsTags.values())]: # temporary fix to include netstim conns allConns.extend([conn for conn in postCell.conns if conn[‘preGid’] == ‘NetStim’])

@@ -321,7 +321,7 @@ connGroup[‘synMech’] = ‘__grouped__’+connGroup[‘synMech’] connsGroup[iConn] = connGroup except:

print ’ Warning: Grouped synMechs %s not found’ % (str(connGroup))
print(’ Warning: Grouped synMechs %s not found’ % (str(connGroup))) else: conns = allConns

@@ -356,7 +356,7 @@ elif subConnParam[‘density’][‘type’] == ‘1Dmap’: # 1D segNumSyn = self._interpolateSegmentSigma(postCell, secList, None, gridY, gridSigma) # move method to Cell!

totSyn = sum([sum(nsyn) for nsyn in segNumSyn.values()]) # summed density
totSyn = sum([sum(nsyn) for nsyn in list(segNumSyn.values())]) # summed density scaleNumSyn = float(len(conns))/float(totSyn) if totSyn>0 else 0.0 diffList = [] for sec in segNumSyn:

@@ -368,7 +368,7 @@ if diff > 0: diffList.append([diff,sec,seg])

totSynRescale = sum([sum(nsyn) for nsyn in segNumSyn.values()])
totSynRescale = sum([sum(nsyn) for nsyn in list(segNumSyn.values())])
if totSynRescale < len(conns):

@@ -389,7 +389,7 @@ subConnParam[‘density’][‘gridValues’] = list(subConnParam[‘density’][‘gridValues’])

newSecs, newLocs = [], []

for sec, nsyns in segNumSyn.iteritems():
for sec, nsyns in segNumSyn.items(): for i, seg in enumerate(postCell.secs[sec][‘hSec’]): for isyn in range(nsyns[i]): newSecs.append(sec)

@@ -401,10 +401,10 @@

if ‘soma’ in postCell.secs: secOrig = ‘soma’

elif any([secName.startswith(‘som’) for secName in postCell.secs.keys()]):
secOrig = next(secName for secName in postCell.secs.keys() if secName.startswith(‘soma’))
elif any([secName.startswith(‘som’) for secName in list(postCell.secs.keys())]):
secOrig = next(secName for secName in list(postCell.secs.keys()) if secName.startswith(‘soma’)) else:
secOrig = postCell.secs.keys()[0]
secOrig = list(postCell.secs.keys())[0]
#print self.fromtodistance(postCell.secs[secOrig](0.5), postCell.secs[‘secs’][conn[‘sec’]](conn[‘loc’]))

@@ -455,7 +455,7 @@ sim.cfg.createNEURONObj = False sim.cfg.addSynMechs = False

for connParamLabel,connParamTemp in self.params.connParams.iteritems(): # for each conn rule or parameter set
for connParamLabel,connParamTemp in self.params.connParams.items(): # for each conn rule or parameter set connParam = connParamTemp.copy() connParam[‘label’] = connParamLabel

@@ -496,12 +496,12 @@ nodeSynapses = sum([len(cell.conns) for cell in sim.net.cells]) nodeConnections = sum([len(set([conn[‘preGid’] for conn in cell.conns])) for cell in sim.net.cells])

print(’ Number of connections on node %i: %i ’ % (sim.rank, nodeConnections))
print((’ Number of connections on node %i: %i ’ % (sim.rank, nodeConnections))) if nodeSynapses != nodeConnections:
print(’ Number of synaptic contacts on node %i: %i ’ % (sim.rank, nodeSynapses))
print((’ Number of synaptic contacts on node %i: %i ’ % (sim.rank, nodeSynapses))) sim.pc.barrier() sim.timing(‘stop’, ‘connectTime’)
if sim.rank == 0 and sim.cfg.timing: print(’ Done; cell connection time = %0.2f s.’ % sim.timingData[‘connectTime’])
if sim.rank == 0 and sim.cfg.timing: print((’ Done; cell connection time = %0.2f s.’ % sim.timingData[‘connectTime’]))
return [cell.conns for cell in self.cells]

@@ -512,17 +512,17 @@ def _findCellsCondition(self, allCellTags, conds): try: cellsTags = dict(allCellTags)

for condKey,condValue in conds.iteritems(): # Find subset of cells that match presyn criteria
for condKey,condValue in conds.items(): # Find subset of cells that match presyn criteria if condKey in [‘x’,’y’,’z’,’xnorm’,’ynorm’,’znorm’]:
cellsTags = {gid: tags for (gid,tags) in cellsTags.iteritems() if condValue[0] <= tags.get(condKey, None) < condValue[1]} # dict with pre cell tags
cellsTags = {gid: tags for (gid,tags) in cellsTags.items() if condValue[0] <= tags.get(condKey, None) < condValue[1]} # dict with pre cell tags prePops = {} else: if isinstance(condValue, list):
cellsTags = {gid: tags for (gid,tags) in cellsTags.iteritems() if tags.get(condKey, None) in condValue} # dict with pre cell tags
prePops = {i: tags for (i,tags) in prePops.iteritems() if (condKey in tags) and (tags.get(condKey, None) in condValue)}
cellsTags = {gid: tags for (gid,tags) in cellsTags.items() if tags.get(condKey, None) in condValue} # dict with pre cell tags
prePops = {i: tags for (i,tags) in prePops.items() if (condKey in tags) and (tags.get(condKey, None) in condValue)} else:
cellsTags = {gid: tags for (gid,tags) in cellsTags.iteritems() if tags.get(condKey, None) == condValue} # dict with pre cell tags
prePops = {i: tags for (i,tags) in prePops.iteritems() if (condKey in tags) and (tags.get(condKey, None) == condValue)}
cellsTags = {gid: tags for (gid,tags) in cellsTags.items() if tags.get(condKey, None) == condValue} # dict with pre cell tags
prePops = {i: tags for (i,tags) in prePops.items() if (condKey in tags) and (tags.get(condKey, None) == condValue)} except: return None

@@ -537,16 +537,16 @@ preCellsTags = dict(allCellTags) # initialize with all presyn cells (make copy) postCellsTags = None

for condKey,condValue in preConds.iteritems(): # Find subset of cells that match presyn criteria
for condKey,condValue in preConds.items(): # Find subset of cells that match presyn criteria if condKey in [‘x’,’y’,’z’,’xnorm’,’ynorm’,’znorm’]:
preCellsTags = {gid: tags for (gid,tags) in preCellsTags.iteritems() if condValue[0] <= tags.get(condKey, None) < condValue[1]} # dict with pre cell tags
preCellsTags = {gid: tags for (gid,tags) in preCellsTags.items() if condValue[0] <= tags.get(condKey, None) < condValue[1]} # dict with pre cell tags #prePops = {} else: if isinstance(condValue, list):
preCellsTags = {gid: tags for (gid,tags) in preCellsTags.iteritems() if tags.get(condKey, None) in condValue} # dict with pre cell tags
preCellsTags = {gid: tags for (gid,tags) in preCellsTags.items() if tags.get(condKey, None) in condValue} # dict with pre cell tags #prePops = {i: tags for (i,tags) in prePops.iteritems() if (condKey in tags) and (tags.get(condKey, None) in condValue)} else:
preCellsTags = {gid: tags for (gid,tags) in preCellsTags.iteritems() if tags.get(condKey, None) == condValue} # dict with pre cell tags
preCellsTags = {gid: tags for (gid,tags) in preCellsTags.items() if tags.get(condKey, None) == condValue} # dict with pre cell tags #prePops = {i: tags for (i,tags) in prePops.iteritems() if (condKey in tags) and (tags.get(condKey, None) == condValue)}

@@ -558,13 +558,13 @@

if preCellsTags: # only check post if there are pre postCellsTags = allCellTags

for condKey,condValue in postConds.iteritems(): # Find subset of cells that match postsyn criteria
for condKey,condValue in postConds.items(): # Find subset of cells that match postsyn criteria if condKey in [‘x’,’y’,’z’,’xnorm’,’ynorm’,’znorm’]:
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.iteritems() if condValue[0] <= tags.get(condKey, None) < condValue[1]} # dict with post Cell objects} # dict with pre cell tags
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.items() if condValue[0] <= tags.get(condKey, None) < condValue[1]} # dict with post Cell objects} # dict with pre cell tags elif isinstance(condValue, list):
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.iteritems() if tags.get(condKey, None) in condValue} # dict with post Cell objects
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.items() if tags.get(condKey, None) in condValue} # dict with post Cell objects else:
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.iteritems() if tags.get(condKey, None) == condValue} # dict with post Cell objects
postCellsTags = {gid: tags for (gid,tags) in postCellsTags.items() if tags.get(condKey, None) == condValue} # dict with post Cell objects #except:

@@ -610,14 +610,14 @@ sqrt(preConds[‘znorm’] - postConds[‘znorm’]))

for k,v in self.params.__dict__.iteritems():
for k,v in self.params.__dict__.items(): if isinstance(v, Number): dictVars[k] = v
for paramStrFunc in paramsStrFunc: strFunc = connParam[paramStrFunc] # string containing function
strVars = [var for var in dictVars.keys() if var in strFunc and var+’norm’ not in strFunc] # get list of variables used (eg. post_ynorm or dist_xyz)
strVars = [var for var in list(dictVars.keys()) if var in strFunc and var+’norm’ not in strFunc] # get list of variables used (eg. post_ynorm or dist_xyz) lambdaStr = ‘lambda ’ + ‘,’.join(strVars) +’: ’ + strFunc # convert to lambda function lambdaFunc = eval(lambdaStr)

@@ -628,19 +628,19 @@

connParam[paramStrFunc+’Func’] = {(preGid,postGid): lambdaFunc( **{strVar: dictVars[strVar] if isinstance(dictVars[strVar], Number) else dictVars[strVar](preCellTags, postCellTags) for strVar in strVars})

for preGid,preCellTags in preCellsTags.iteritems() for postGid,postCellTags in postCellsTags.iteritems()}
for preGid,preCellTags in preCellsTags.items() for postGid,postCellTags in postCellsTags.items()}
elif paramStrFunc in [‘convergence’]:

connParam[paramStrFunc+’Func’] = {postGid: lambdaFunc( **{strVar: dictVars[strVar] if isinstance(dictVars[strVar], Number) else dictVars[strVar](None, postCellTags) for strVar in strVars})
for postGid,postCellTags in postCellsTags.iteritems()}
for postGid,postCellTags in postCellsTags.items()}
elif paramStrFunc in [‘divergence’]:

connParam[paramStrFunc+’Func’] = {preGid: lambdaFunc( **{strVar: dictVars[strVar] if isinstance(dictVars[strVar], Number) else dictVars[strVar](preCellTags, None) for strVar in strVars})
for preGid, preCellTags in preCellsTags.iteritems()}
for preGid, preCellTags in preCellsTags.items()}
else:

@@ -653,20 +653,20 @@ ############################################################################### def fullConn (self, preCellsTags, postCellsTags, connParam): ”’ Generates connections between all pre and post-syn cells ”’

if sim.cfg.verbose: print ‘Generating set of all-to-all connections (rule: %s) …’ % (connParam[‘label’])
if sim.cfg.verbose: print(‘Generating set of all-to-all connections (rule: %s) …’ % (connParam[‘label’]))
paramsStrFunc = [param for param in [p+’Func’ for p in self.connStringFuncParams] if param in connParam]

for paramStrFunc in paramsStrFunc:
seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+preCellsTags.keys()[0]+postCellsTags.keys()[0])))
connParam[paramStrFunc[:-4]+’List’] = {(preGid,postGid): connParam[paramStrFunc](**{k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].iteritems()})
for preGid,preCellTags in preCellsTags.iteritems() for postGid,postCellTags in postCellsTags.iteritems()}
seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+list(preCellsTags.keys())[0]+list(postCellsTags.keys())[0])))
connParam[paramStrFunc[:-4]+’List’] = {(preGid,postGid): connParam[paramStrFunc](**{k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].items()})
for preGid,preCellTags in preCellsTags.items() for postGid,postCellTags in postCellsTags.items()}
for postCellGid in postCellsTags: # for each postsyn cell if postCellGid in self.lid2gid: # check if postsyn is in this node’s list of gids
for preCellGid, preCellTags in preCellsTags.iteritems(): # for each presyn cell
for preCellGid, preCellTags in preCellsTags.items(): # for each presyn cell self._addCellConn(connParam, preCellGid, postCellGid) # add connection

@@ -675,21 +675,21 @@ ############################################################################### def probConn (self, preCellsTags, postCellsTags, connParam): ”’ Generates connections between all pre and post-syn cells based on probability values”’

if sim.cfg.verbose: print ‘Generating set of probabilistic connections (rule: %s) …’ % (connParam[‘label’])

-

seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+preCellsTags.keys()[-1]+postCellsTags.keys()[-1])))
if sim.cfg.verbose: print(‘Generating set of probabilistic connections (rule: %s) …’ % (connParam[‘label’]))

+

seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+list(preCellsTags.keys())[-1]+list(postCellsTags.keys())[-1]))) allRands = {(preGid,postGid): random() for preGid in preCellsTags for postGid in postCellsTags} # Create an array of random numbers for checking each connection
paramsStrFunc = [param for param in [p+’Func’ for p in self.connStringFuncParams] if param in connParam]
for postCellGid,postCellTags in postCellsTags.iteritems(): # for each postsyn cell
for postCellGid,postCellTags in postCellsTags.items(): # for each postsyn cell if postCellGid in self.lid2gid: # check if postsyn is in this node
for preCellGid, preCellTags in preCellsTags.iteritems(): # for each presyn cell
for preCellGid, preCellTags in preCellsTags.items(): # for each presyn cell probability = connParam[‘probabilityFunc’][preCellGid,postCellGid] if ‘probabilityFunc’ in connParam else connParam[‘probability’]
for paramStrFunc in paramsStrFunc: # call lambda functions to get weight func args
connParam[paramStrFunc+’Args’] = {k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].iteritems()}
connParam[paramStrFunc+’Args’] = {k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].items()}
if probability >= allRands[preCellGid,postCellGid]: seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+postCellGid+preCellGid)))

@@ -701,22 +701,22 @@ ############################################################################### def convConn (self, preCellsTags, postCellsTags, connParam): ”’ Generates connections between all pre and post-syn cells based on probability values”’

if sim.cfg.verbose: print ‘Generating set of convergent connections (rule: %s) …’ % (connParam[‘label’])
if sim.cfg.verbose: print(‘Generating set of convergent connections (rule: %s) …’ % (connParam[‘label’]))
paramsStrFunc = [param for param in [p+’Func’ for p in self.connStringFuncParams] if param in connParam]
for postCellGid,postCellTags in postCellsTags.iteritems(): # for each postsyn cell
for postCellGid,postCellTags in postCellsTags.items(): # for each postsyn cell if postCellGid in self.lid2gid: # check if postsyn is in this node convergence = connParam[‘convergenceFunc’][postCellGid] if ‘convergenceFunc’ in connParam else connParam[‘convergence’] # num of presyn conns / postsyn cell convergence = max(min(int(round(convergence)), len(preCellsTags)), 0) seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+postCellGid)))
preCellsSample = sample(preCellsTags.keys(), convergence) # selected gids of presyn cells
preCellsConv = {k:v for k,v in preCellsTags.iteritems() if k in preCellsSample} # dict of selected presyn cells tags
for preCellGid, preCellTags in preCellsConv.iteritems(): # for each presyn cell
preCellsSample = sample(list(preCellsTags.keys()), convergence) # selected gids of presyn cells
preCellsConv = {k:v for k,v in preCellsTags.items() if k in preCellsSample} # dict of selected presyn cells tags
for preCellGid, preCellTags in preCellsConv.items(): # for each presyn cell
for paramStrFunc in paramsStrFunc: # call lambda functions to get weight func args
connParam[paramStrFunc+’Args’] = {k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].iteritems()}
connParam[paramStrFunc+’Args’] = {k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].items()}
seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+postCellGid+preCellGid))) if preCellGid != postCellGid: # if not self-connection

@@ -728,21 +728,21 @@ ############################################################################### def divConn (self, preCellsTags, postCellsTags, connParam): ”’ Generates connections between all pre and post-syn cells based on probability values”’

if sim.cfg.verbose: print ‘Generating set of divergent connections (rule: %s) …’ % (connParam[‘label’])
if sim.cfg.verbose: print(‘Generating set of divergent connections (rule: %s) …’ % (connParam[‘label’]))
paramsStrFunc = [param for param in [p+’Func’ for p in self.connStringFuncParams] if param in connParam]
for preCellGid, preCellTags in preCellsTags.iteritems(): # for each presyn cell
for preCellGid, preCellTags in preCellsTags.items(): # for each presyn cell divergence = connParam[‘divergenceFunc’][preCellGid] if ‘divergenceFunc’ in connParam else connParam[‘divergence’] # num of presyn conns / postsyn cell divergence = max(min(int(round(divergence)), len(postCellsTags)), 0) seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+preCellGid))) postCellsSample = sample(postCellsTags, divergence) # selected gids of postsyn cells
postCellsDiv = {postGid:postConds for postGid,postConds in postCellsTags.iteritems() if postGid in postCellsSample and postGid in self.lid2gid} # dict of selected postsyn cells tags
for postCellGid, postCellTags in postCellsDiv.iteritems(): # for each postsyn cell
postCellsDiv = {postGid:postConds for postGid,postConds in postCellsTags.items() if postGid in postCellsSample and postGid in self.lid2gid} # dict of selected postsyn cells tags
for postCellGid, postCellTags in postCellsDiv.items(): # for each postsyn cell
for paramStrFunc in paramsStrFunc: # call lambda functions to get weight func args
connParam[paramStrFunc+’Args’] = {k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].iteritems()}
connParam[paramStrFunc+’Args’] = {k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].items()}
seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+postCellGid+preCellGid))) if preCellGid != postCellGid: # if not self-connection

@@ -754,15 +754,15 @@ ############################################################################### def fromListConn (self, preCellsTags, postCellsTags, connParam): ”’ Generates connections between all pre and post-syn cells based list of relative cell ids”’

if sim.cfg.verbose: print ‘Generating set of connections from list (rule: %s) …’ % (connParam[‘label’])
if sim.cfg.verbose: print(‘Generating set of connections from list (rule: %s) …’ % (connParam[‘label’]))
paramsStrFunc = [param for param in [p+’Func’ for p in self.connStringFuncParams] if param in connParam] for paramStrFunc in paramsStrFunc:
seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+preCellsTags.keys()[0]+postCellsTags.keys()[0])))
connParam[paramStrFunc[:-4]+’List’] = {(preGid,postGid): connParam[paramStrFunc](**{k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].iteritems()})
for preGid,preCellTags in preCellsTags.iteritems() for postGid,postCellTags in postCellsTags.iteritems()}
seed(sim.id32(‘%d’%(sim.cfg.seeds[‘conn’]+list(preCellsTags.keys())[0]+list(postCellsTags.keys())[0])))
connParam[paramStrFunc[:-4]+’List’] = {(preGid,postGid): connParam[paramStrFunc](**{k:v if isinstance(v, Number) else v(preCellTags,postCellTags) for k,v in connParam[paramStrFunc+’Vars’].items()})
for preGid,preCellTags in preCellsTags.items() for postGid,postCellTags in postCellsTags.items()}
if isinstance(connParam[‘weight’], list): connParam[‘weightFromList’] = list(connParam[‘weight’]) # if weight is a list, copy to weightFromList if isinstance(connParam[‘delay’], list): connParam[‘delayFromList’] = list(connParam[‘delay’]) # if delay is a list, copy to delayFromList

@@ -851,7 +851,7 @@ sim._gatherCells() # update allCells

sim.timing(‘stop’, ‘modifyCellsTime’)

if sim.rank == 0 and sim.cfg.timing: print(’ Done; cells modification time = %0.2f s.’ % sim.timingData[‘modifyCellsTime’])
if sim.rank == 0 and sim.cfg.timing: print((’ Done; cells modification time = %0.2f s.’ % sim.timingData[‘modifyCellsTime’]))

############################################################################### @@ -870,7 +870,7 @@ sim._gatherCells() # update allCells

sim.timing(‘stop’, ‘modifySynMechsTime’)

if sim.rank == 0 and sim.cfg.timing: print(’ Done; syn mechs modification time = %0.2f s.’ % sim.timingData[‘modifySynMechsTime’])
if sim.rank == 0 and sim.cfg.timing: print((’ Done; syn mechs modification time = %0.2f s.’ % sim.timingData[‘modifySynMechsTime’]))

@@ -890,7 +890,7 @@ sim._gatherCells() # update allCells

sim.timing(‘stop’, ‘modifyConnsTime’)

if sim.rank == 0 and sim.cfg.timing: print(’ Done; connections modification time = %0.2f s.’ % sim.timingData[‘modifyConnsTime’])
if sim.rank == 0 and sim.cfg.timing: print((’ Done; connections modification time = %0.2f s.’ % sim.timingData[‘modifyConnsTime’]))

############################################################################### @@ -909,9 +909,9 @@ sim._gatherCells() # update allCells

sim.timing(‘stop’, ‘modifyStimsTime’)

if sim.rank == 0 and sim.cfg.timing: print(’ Done; stims modification time = %0.2f s.’ % sim.timingData[‘modifyStimsTime’])

- - - - -

if sim.rank == 0 and sim.cfg.timing: print((’ Done; stims modification time = %0.2f s.’ % sim.timingData[‘modifyStimsTime’]))

+ + + + + RefactoringTool: Writing converted netpyne/network.py to netpyne_py3/network.py. RefactoringTool: Refactored netpyne/neuromlFuncs.py — netpyne/neuromlFuncs.py (original) +++ netpyne/neuromlFuncs.py (refactored) @@ -17,7 +17,7 @@ import pprint; pp = pprint.PrettyPrinter(depth=6) import math from collections import OrderedDict -import sim, specs +from . import sim, specs

############################################################################### ### Get connection centric network representation as used in NeuroML2 @@ -26,8 +26,8 @@

nn = {}

for np_pop in net.pops.values():
print(“Adding conns for: %s”%np_pop.tags)
for np_pop in list(net.pops.values()):
print((“Adding conns for: %s”%np_pop.tags)) if not np_pop.tags[‘cellModel’] == ‘NetStim’: for cell in net.cells: if cell.gid in np_pop.cellGids:

@@ -44,10 +44,10 @@ synMech = conn[‘synMech’] threshold = conn[‘threshold’]

if sim.cfg.verbose: print(” Conn %s[%i]->%s[%i] with %s, w: %s, d: %s”%(popPre, indexPre,popPost, indexPost, synMech, weight, delay))
if sim.cfg.verbose: print((” Conn %s[%i]->%s[%i] with %s, w: %s, d: %s”%(popPre, indexPre,popPost, indexPost, synMech, weight, delay)))
projection_info = (popPre,popPost,synMech)
if not projection_info in nn.keys():
if not projection_info in list(nn.keys()): nn[projection_info] = []
nn[projection_info].append({‘indexPre’:indexPre,’indexPost’:indexPost,’weight’:weight,’delay’:delay})

@@ -65,9 +65,9 @@

stims = {}

for np_pop in net.pops.values():
for np_pop in list(net.pops.values()): if not np_pop.tags[‘cellModel’] == ‘NetStim’:
print(“Adding stims for: %s”%np_pop.tags)
print((“Adding stims for: %s”%np_pop.tags)) for cell in net.cells: if cell.gid in np_pop.cellGids: pop, index = gids_vs_pop_indices[cell.gid]

@@ -91,7 +91,7 @@ name_stim = ‘NetStim_%s_%s_%s_%s_%s’%(ref,pop,rate,noise,synMech)

stim_info = (name_stim, pop, rate, noise,synMech)

if not stim_info in stims.keys():
if not stim_info in list(stims.keys()): stims[stim_info] = []

@@ -109,9 +109,9 @@ def _export_synapses (net, nml_doc):

syn_types = {}

for id,syn in net.params.synMechParams.iteritems():
for id,syn in net.params.synMechParams.items(): syn_types[id]=syn[‘mod’]
if sim.cfg.verbose: print(‘Exporting details of syn: %s’%syn)
if sim.cfg.verbose: print((‘Exporting details of syn: %s’%syn)) if syn[‘mod’] == ‘Exp2Syn’: syn0 = neuroml.ExpTwoSynapse(id=id, gbase=’1uS’,

@@ -188,7 +188,7 @@

net = sim.net

print(“Exporting network to NeuroML 2, reference: %s”%reference)
print((“Exporting network to NeuroML 2, reference: %s”%reference))
import neuroml import neuroml.writers as writers

@@ -205,30 +205,30 @@

cells_added = []

for np_pop in net.pops.values():
if sim.cfg.verbose: print(“– Adding population: %s”%np_pop.tags)
for np_pop in list(net.pops.values()):
if sim.cfg.verbose: print((“– Adding population: %s”%np_pop.tags))
cell_param_set = {}
for cell_name in net.params.cellParams.keys():
for cell_name in list(net.params.cellParams.keys()): cell_param_set0 = net.params.cellParams[cell_name] someMatches = False someMisMatches = False
if sim.cfg.verbose: print(” – Comparing conds: %s”%cell_param_set0)
if sim.cfg.verbose: print((” – Comparing conds: %s”%cell_param_set0)) for cond in cell_param_set0[‘conds’]: if len(cell_param_set0[‘conds’][cond])>0: if cond in np_pop.tags and cell_param_set0[‘conds’][cond] == np_pop.tags[cond]:
if sim.cfg.verbose: print ” Cond: %s matches…”%cond
if sim.cfg.verbose: print(” Cond: %s matches…”%cond) someMatches = True else:
if sim.cfg.verbose: print ” Cond: %s DOESN’T match (%s != %s)…”%(cond,cell_param_set0[‘conds’][cond],np_pop.tags[cond] if cond in np_pop.tags else “???”)
if sim.cfg.verbose: print(” Cond: %s DOESN’T match (%s != %s)…”%(cond,cell_param_set0[‘conds’][cond],np_pop.tags[cond] if cond in np_pop.tags else “???”)) someMisMatches = True if someMatches and not someMisMatches:
if sim.cfg.verbose: print(” Matches: %s”%cell_param_set0)
if sim.cfg.verbose: print((” Matches: %s”%cell_param_set0)) cell_param_set.update(cell_param_set0)
if not np_pop.tags[‘cellModel’] == ‘NetStim’ and len(cell_param_set)==0:
print(“Error, could not find cellParams for %s”%np_pop.tags)
print((“Error, could not find cellParams for %s”%np_pop.tags)) exit(-1)
if not np_pop.tags[‘cellModel’] == ‘NetStim’:

@@ -243,7 +243,7 @@

if not np_pop.tags[‘cellModel’] == ‘NetStim’ and not str(cell_param_set) in cells_added:

if sim.cfg.verbose: print(“--------------- Adding a cell from pop %s: \n%s”%(np_pop.tags,cell_param_set))
if sim.cfg.verbose: print((“--------------- Adding a cell from pop %s: \n%s”%(np_pop.tags,cell_param_set)))

@@ -252,12 +252,12 @@

if len(cell_param_set[‘secs’]) == 1 \ and soma is not None\

and not soma.has_key(‘mechs’) \
and ‘mechs’ not in soma \ and len(soma[‘pointps’]) == 1:
pproc = soma[‘pointps’].values()[0]

-

print(“Assuming abstract cell with behaviour set by single point process: %s!”%pproc)
pproc = list(soma[‘pointps’].values())[0]

+

print((“Assuming abstract cell with behaviour set by single point process: %s!”%pproc))
if pproc[‘mod’] == ‘Izhi2007b’: izh = neuroml.Izhikevich2007Cell(id=cell_id)

@@ -275,7 +275,7 @@

nml_doc.izhikevich2007_cells.append(izh) else:

print(“Unknown point process: %s; can’t convert to NeuroML 2 equivalent!”%pproc[‘mod’])
print((“Unknown point process: %s; can’t convert to NeuroML 2 equivalent!”%pproc[‘mod’])) exit(1) else: print(“Assuming normal cell with behaviour set by ion channel mechanisms!”)

@@ -299,7 +299,7 @@

parentDistal = neuroml.Point3DWithDiam(x=0,y=0,z=0,diameter=0)

for np_sec_name in cell_param_set[‘secs’].keys():
for np_sec_name in list(cell_param_set[‘secs’].keys()):
parent_seg = None

@@ -307,7 +307,7 @@ nml_seg = neuroml.Segment(id=count,name=np_sec_name) nml_segs[np_sec_name] = nml_seg

if np_sec.has_key(‘topol’) and len(np_sec[‘topol’])>0:
if ‘topol’ in np_sec and len(np_sec[‘topol’])>0: parent_seg = nml_segs[np_sec[‘topol’][‘parentSec’]] nml_seg.parent = neuroml.SegmentParent(segments=parent_seg.id)

@@ -315,11 +315,11 @@ print(“Currently only support cell topol with parentX == 1.0 and childX == 0”) exit(1)

if not ( (not np_sec[‘geom’].has_key(‘pt3d’)) or len(np_sec[‘geom’][‘pt3d’])==0 or len(np_sec[‘geom’][‘pt3d’])==2 ):
print(“Currently only support cell geoms with 2 pt3ds (or 0 and diam/L specified): %s”%np_sec[‘geom’])
if not ( (‘pt3d’ not in np_sec[‘geom’]) or len(np_sec[‘geom’][‘pt3d’])==0 or len(np_sec[‘geom’][‘pt3d’])==2 ):
print((“Currently only support cell geoms with 2 pt3ds (or 0 and diam/L specified): %s”%np_sec[‘geom’])) exit(1)
if (not np_sec[‘geom’].has_key(‘pt3d’) or len(np_sec[‘geom’][‘pt3d’])==0):
if (‘pt3d’ not in np_sec[‘geom’] or len(np_sec[‘geom’][‘pt3d’])==0):
if parent_seg == None: nml_seg.proximal = neuroml.Point3DWithDiam(x=parentDistal.x,

@@ -360,7 +360,7 @@ ”’ See suny-downstate-medical-center#130 ”’

cm = np_sec[‘geom’][‘cm’] if np_sec[‘geom’].has_key(‘cm’) else 1
cm = np_sec[‘geom’][‘cm’] if ‘cm’ in np_sec[‘geom’] else 1 if isinstance(cm,dict) and len(cm)==0: cm = 1 mp.specific_capacitances.append(neuroml.SpecificCapacitance(value=”%s uF_per_cm2”%cm,

@@ -371,7 +371,7 @@

mp.spike_threshes.append(neuroml.SpikeThresh(value=”%s mV”%sim.net.params.defaultThreshold))

for mech_name in np_sec[‘mechs’].keys():
for mech_name in list(np_sec[‘mechs’].keys()): mech = np_sec[‘mechs’][mech_name] if mech_name == ‘hh’:

@@ -418,7 +418,7 @@ ion=’non_specific’) mp.channel_densities.append(leak_cd) else:

print(“Currently NML2 export only supports mech hh, not: %s”%mech_name)
print((“Currently NML2 export only supports mech hh, not: %s”%mech_name)) exit(1)

@@ -427,9 +427,9 @@ cells_added.append(str(cell_param_set))

for np_pop in net.pops.values():
for np_pop in list(net.pops.values()): index = 0
print(“Adding population: %s”%np_pop.tags)
print((“Adding population: %s”%np_pop.tags))
type = ‘populationList’ if not np_pop.tags[‘cellModel’] == ‘NetStim’:

@@ -453,11 +453,11 @@ nn = _convertNetworkRepresentation(net, gids_vs_pop_indices)

half_elect_conns_added = []

for proj_info in nn.keys():
for proj_info in list(nn.keys()):
prefix = “NetConn” popPre,popPost,synMech = proj_info
if sim.cfg.verbose: print(“Adding proj: %s->%s (%s)”%(popPre,popPost,synMech))
if sim.cfg.verbose: print((“Adding proj: %s->%s (%s)”%(popPre,popPost,synMech)))
if syn_types[synMech]!=’ElectSyn’: projection = neuroml.Projection(id=”%s_%s_%s_%s”%(prefix,popPre, popPost,synMech),

@@ -478,7 +478,7 @@

for conn in nn[proj_info]:

if sim.cfg.verbose: print(“Adding conn %s”%conn)
if sim.cfg.verbose: print((“Adding conn %s”%conn))
if syn_types[synMech]!=’ElectSyn’: connection = neuroml.ConnectionWD(id=index, \

@@ -535,10 +535,10 @@ if stimulations: stims = _convertStimulationRepresentation(net, gids_vs_pop_indices, nml_doc)

for stim_info in stims.keys():
for stim_info in list(stims.keys()): name_stim, post_pop, rate, noise, synMech = stim_info
if sim.cfg.verbose: print(“Adding stim: %s”%[stim_info])
if sim.cfg.verbose: print((“Adding stim: %s”%[stim_info]))
if noise==0: source = neuroml.SpikeGenerator(id=name_stim,period=”%ss”%(1./rate))

@@ -596,10 +596,10 @@ copy_neuroml = False, include_extra_files = [], gen_plots_for_all_v = False,

gen_plots_for_only_populations = populations_vs_components.keys(),
gen_plots_for_only_populations = list(populations_vs_components.keys()), gen_saves_for_all_v = False, plot_all_segments = False,
gen_saves_for_only_populations = populations_vs_components.keys(),
gen_saves_for_only_populations = list(populations_vs_components.keys()), save_all_segments = False, seed=1234)

@@ -645,18 +645,18 @@

def finalise(self):

for popParam in self.popParams.keys():
for popParam in list(self.popParams.keys()): self.netParams.popParams[popParam] = self.popParams[popParam]
for cellParam in self.cellParams.keys():
for cellParam in list(self.cellParams.keys()): self.netParams.cellParams[cellParam] = self.cellParams[cellParam]
for proj_id in self.projection_infos.keys():
for proj_id in list(self.projection_infos.keys()): projName, prePop, postPop, synapse, ptype = self.projection_infos[proj_id]
self.netParams.synMechParams[synapse] = {‘mod’: synapse}
for stimName in self.stimSources.keys():
for stimName in list(self.stimSources.keys()): self.netParams.stimSourceParams[stimName] = self.stimSources[stimName] self.netParams.stimTargetParams[stimName] = self.stimLists[stimName]

@@ -683,7 +683,7 @@

def handleNetwork(self, network_id, notes, temperature=None): if temperature:

print(“\n****************\n Need to set temperature to %s!!!\n********************”%temperature)
print((“\n****************\n Need to set temperature to %s!!!\n********************”%temperature))

# @@ -691,7 +691,7 @@

def handlePopulation(self, population_id, component, size, component_obj):

if self.verbose: print(“A population: %s with %i of %s (%s)”%(population_id,size,component,component_obj))
if self.verbose: print((“A population: %s with %i of %s (%s)”%(population_id,size,component,component_obj)))
self.pop_ids_vs_components[population_id] = component_obj

@@ -755,11 +755,11 @@

else:

ordered_segs, cumulative_lengths = cell.get_ordered_segments_in_groups(cellRule[‘secs’].keys(),include_cumulative_lengths=True)
ordered_segs, cumulative_lengths = cell.get_ordered_segments_in_groups(list(cellRule[‘secs’].keys()),include_cumulative_lengths=True) self.pop_ids_vs_ordered_segs[population_id] = ordered_segs self.pop_ids_vs_cumulative_lengths[population_id] = cumulative_lengths
for section in cellRule[‘secs’].keys():
for section in list(cellRule[‘secs’].keys()): #print(“ggg %s: %s”%(section,ordered_segs[section])) for seg in ordered_segs[section]: prox, dist = self._get_prox_dist(seg, seg_ids_vs_segs)

@@ -770,7 +770,7 @@ parent_seg = seg_ids_vs_segs[seg.parent.segments] parent_sec = None #TODO: optimise

for sec in ordered_segs.keys():
for sec in list(ordered_segs.keys()): if parent_seg.id in [s.id for s in ordered_segs[sec]]: parent_sec = sec fract = float(seg.parent.fraction_along)

@@ -797,7 +797,7 @@ seg_grps_vs_nrn_sections[seg_grp.id].append(section_name) else: seg_grps_vs_nrn_sections[seg_grp.id].append(inc.segment_groups)

if not cellRule[‘secLists’].has_key(seg_grp.id): cellRule[‘secLists’][seg_grp.id] = []
if seg_grp.id not in cellRule[‘secLists’]: cellRule[‘secLists’][seg_grp.id] = [] cellRule[‘secLists’][seg_grp.id].append(inc.segment_groups)
if not seg_grp.neuro_lex_id or seg_grp.neuro_lex_id !=”sao864921383”:

@@ -819,7 +819,7 @@ if cm.ion and cm.ion == ‘non_specific’: mech[‘e’] = erev else:

if not cellRule[‘secs’][section_name][‘ions’].has_key(cm.ion):
if cm.ion not in cellRule[‘secs’][section_name][‘ions’]: cellRule[‘secs’][section_name][‘ions’][cm.ion] = {} cellRule[‘secs’][section_name][‘ions’][cm.ion][‘e’] = erev

@@ -840,7 +840,7 @@ pass ##mech[‘e’] = erev else:

if not cellRule[‘secs’][section_name][‘ions’].has_key(cm.ion):
if cm.ion not in cellRule[‘secs’][section_name][‘ions’]: cellRule[‘secs’][section_name][‘ions’][cm.ion] = {} ##cellRule[‘secs’][section_name][‘ions’][cm.ion][‘e’] = erev

@@ -899,7 +899,7 @@

popInfo[‘cellType’] = component

if self.verbose: print(“Abstract cell: %s”%(isinstance(component_obj,BaseCell)))
if self.verbose: print((“Abstract cell: %s”%(isinstance(component_obj,BaseCell))))
if not isinstance(component_obj,BaseCell): popInfo[‘originalFormat’] = ‘NeuroML2_SpikeSource’

@@ -943,7 +943,7 @@

def _convert_to_nrn_section_location(self, population_id, seg_id, fract_along):

if not self.pop_ids_vs_seg_ids_vs_segs.has_key(population_id) or not self.pop_ids_vs_seg_ids_vs_segs[population_id].has_key(seg_id):
if population_id not in self.pop_ids_vs_seg_ids_vs_segs or seg_id not in self.pop_ids_vs_seg_ids_vs_segs[population_id]: return ‘soma’, 0.5
if not self.pop_ids_vs_use_segment_groups_for_neuron[population_id]:

@@ -951,7 +951,7 @@ return self.pop_ids_vs_seg_ids_vs_segs[population_id][seg_id].name, fract_along else: fract_sec = -1

for sec in self.pop_ids_vs_ordered_segs[population_id].keys():
for sec in list(self.pop_ids_vs_ordered_segs[population_id].keys()): ind = 0 for seg in self.pop_ids_vs_ordered_segs[population_id][sec]: if seg.id == seg_id:

@@ -987,7 +987,7 @@ # def handleProjection(self, projName, prePop, postPop, synapse, hasWeights=False, hasDelays=False, type=”projection”, synapse_obj=None, pre_synapse_obj=None):

if self.verbose: print(“A projection: %s (%s) from %s -> %s with syn: %s” % (projName, type, prePop, postPop, synapse))
if self.verbose: print((“A projection: %s (%s) from %s -> %s with syn: %s” % (projName, type, prePop, postPop, synapse))) self.projection_infos[projName] = (projName, prePop, postPop, synapse, type) self.connections[projName] = []

@@ -1076,7 +1076,7 @@ ‘conds’: {‘pop’:pop_id, ‘cellList’: [cellId]}}

if self.verbose: print(“Input: %s[%s] on %s, cellId: %i, seg: %i (nrn: %s), fract: %f (nrn: %f); ref: %s” % (inputListId,id,pop_id,cellId,segId,nrn_sec,fract,nrn_fract,stimId))
if self.verbose: print((“Input: %s[%s] on %s, cellId: %i, seg: %i (nrn: %s), fract: %f (nrn: %f); ref: %s” % (inputListId,id,pop_id,cellId,segId,nrn_sec,fract,nrn_fract,stimId)))

@@ -1094,7 +1094,7 @@ import pprint pp = pprint.PrettyPrinter(indent=4)

print(“Importing NeuroML 2 network from: %s”%fileName)
print((“Importing NeuroML 2 network from: %s”%fileName))
nmlHandler = None

@@ -1115,7 +1115,7 @@

nmlHandler.finalise()

print(‘Finished import of NeuroML2; populations vs gids NML has calculated: %s’%nmlHandler.gids)
print((‘Finished import of NeuroML2; populations vs gids NML has calculated: %s’%nmlHandler.gids)) #print(‘Connections: %s’%nmlHandler.connections)
if fileName.endswith(“.h5”):

@@ -1133,7 +1133,7 @@

nmlHandler.finalise()

print(‘Finished import: %s’%nmlHandler.gids)
print((‘Finished import: %s’%nmlHandler.gids)) #print(‘Connections: %s’%nmlHandler.connections)

@@ -1146,14 +1146,14 @@

for popLabel,pop in sim.net.pops.iteritems():
if sim.cfg.verbose: print(“gid: %s: %s, %s”%(popLabel,pop, pop.cellGids))
for popLabel,pop in sim.net.pops.items():
if sim.cfg.verbose: print((“gid: %s: %s, %s”%(popLabel,pop, pop.cellGids))) for gid in pop.cellGids: assert gid in nmlHandler.gids[popLabel]
for proj_id in nmlHandler.projection_infos.keys():
for proj_id in list(nmlHandler.projection_infos.keys()): projName, prePop, postPop, synapse, ptype = nmlHandler.projection_infos[proj_id]
if sim.cfg.verbose: print(“Creating connections for %s (%s): %s->%s via %s”%(projName, ptype, prePop, postPop, synapse))
if sim.cfg.verbose: print((“Creating connections for %s (%s): %s->%s via %s”%(projName, ptype, prePop, postPop, synapse)))
preComp = nmlHandler.pop_ids_vs_components[prePop]

@@ -1210,7 +1210,7 @@ cell = sim.net.cells[sim.net.gid2lid[preGapParams[‘gid’]]] cell.addConn(preGapParams)

print(’ Number of connections on node %i: %i ’ % (sim.rank, sum([len(cell.conns) for cell in sim.net.cells])))
print((’ Number of connections on node %i: %i ’ % (sim.rank, sum([len(cell.conns) for cell in sim.net.cells]))))

RefactoringTool: Writing converted netpyne/neuromlFuncs.py to netpyne_py3/neuromlFuncs.py. RefactoringTool: Refactored netpyne/pop.py — netpyne/pop.py (original) +++ netpyne/pop.py (refactored) @@ -10,7 +10,7 @@ from matplotlib.pylab import arange, seed, rand, array, pi, sqrt, sin, cos, arccos import numpy as np from neuron import h # Import NEURON -import sim +from . import sim

############################################################################### @@ -42,7 +42,7 @@ sim.nextHost=0

if sim.cfg.verbose:

print(“Distributed population of %i cells on %s hosts: %s, next: %s”%(numCellsPop,sim.nhosts,hostCells,sim.nextHost))
print((“Distributed population of %i cells on %s hosts: %s, next: %s”%(numCellsPop,sim.nhosts,hostCells,sim.nextHost))) return hostCells

@@ -67,7 +67,7 @@

else: self.tags[‘numCells’] = 1

print ‘Warninig: number or density of cells not specified for population %s; defaulting to numCells = 1’ % (self.tags[‘pop’])
print(‘Warninig: number or density of cells not specified for population %s; defaulting to numCells = 1’ % (self.tags[‘pop’])) cells = self.createCellsFixedNum()
return cells

@@ -113,7 +113,7 @@ for i in self._distributeCells(int(sim.net.params.scale * self.tags[‘numCells’]))[sim.rank]: gid = sim.net.lastGid+i self.cellGids.append(gid) # add gid list of cells belonging to this population - not needed?

cellTags = {k: v for (k, v) in self.tags.iteritems() if k in sim.net.params.popTagsCopiedToCells} # copy all pop tags to cell tags, except those that are pop-specific
cellTags = {k: v for (k, v) in self.tags.items() if k in sim.net.params.popTagsCopiedToCells} # copy all pop tags to cell tags, except those that are pop-specific cellTags[‘pop’] = self.tags[‘pop’] cellTags[‘xnorm’] = randLocs[i,0] # set x location (um) cellTags[‘ynorm’] = randLocs[i,1] # set y location (um)

@@ -122,7 +122,7 @@ cellTags[‘y’] = sim.net.params.sizeY * randLocs[i,1] # set y location (um) cellTags[‘z’] = sim.net.params.sizeZ * randLocs[i,2] # set z location (um) cells.append(self.cellModelClass(gid, cellTags)) # instantiate Cell object

if sim.cfg.verbose: print(‘Cell %d/%d (gid=%d) of pop %s, on node %d, ‘%(i, sim.net.params.scale * self.tags[‘numCells’]-1, gid, self.tags[‘pop’], sim.rank))
if sim.cfg.verbose: print((‘Cell %d/%d (gid=%d) of pop %s, on node %d, ‘%(i, sim.net.params.scale * self.tags[‘numCells’]-1, gid, self.tags[‘pop’], sim.rank))) sim.net.lastGid = sim.net.lastGid + self.tags[‘numCells’] return cells

@@ -157,7 +157,7 @@ strFunc = self.tags[‘density’] # string containing function strVars = [var for var in [‘xnorm’, ‘ynorm’, ‘znorm’] if var in strFunc] # get list of variables used if not len(strVars) == 1:

print ‘Error: density function (%s) for population %s does not include “xnorm”, “ynorm” or “znorm”’%(strFunc,self.tags[‘pop’])
print(‘Error: density function (%s) for population %s does not include “xnorm”, “ynorm” or “znorm”’%(strFunc,self.tags[‘pop’])) return coordFunc = strVars[0] lambdaStr = ‘lambda ’ + coordFunc +’: ’ + strFunc # convert to lambda function

@@ -166,20 ~~166,20 @@ maxRange = self.tags[coordFunc~~‘Range’][1]

interval = 0.001 # interval of location values to evaluate func in order to find the max cell density

maxDensity = max(map(densityFunc, (arange(minRange, maxRange, interval)))) # max cell density
maxDensity = max(list(map(densityFunc, (arange(minRange, maxRange, interval))))) # max cell density maxCells = volume * maxDensity # max number of cells based on max value of density func
seed(sim.id32(‘%d’ % sim.cfg.seeds[‘loc’]+sim.net.lastGid)) # reset random number generator locsAll = minRange + ((maxRange-minRange)) * rand(int(maxCells), 1) # random location values
locsProb = array(map(densityFunc, locsAll)) / maxDensity # calculate normalized density for each location value (used to prune)
locsProb = array(list(map(densityFunc, locsAll))) / maxDensity # calculate normalized density for each location value (used to prune) allrands = rand(len(locsProb)) # create an array of random numbers for checking each location pos
makethiscell = locsProb>allrands # perform test to see whether or not this cell should be included (pruning based on density func) funcLocs = [locsAll[i] for i in range(len(locsAll)) if i in array(makethiscell.nonzero()[0],dtype=’int’)] # keep only subset of yfuncLocs based on density func self.tags[‘numCells’] = len(funcLocs) # final number of cells after pruning of location values based on density func
if sim.cfg.verbose: print ‘Volume=%.2f, maxDensity=%.2f, maxCells=%.0f, numCells=%.0f’%(volume, maxDensity, maxCells, self.tags[‘numCells’])
if sim.cfg.verbose: print(‘Volume=%.2f, maxDensity=%.2f, maxCells=%.0f, numCells=%.0f’%(volume, maxDensity, maxCells, self.tags[‘numCells’])) else:
print ‘Error: Density functions are only implemented for cuboid shaped networks’
print(‘Error: Density functions are only implemented for cuboid shaped networks’) exit(0) else: # NO ynorm-dep self.tags[‘numCells’] = int(self.tags[‘density’] * volume) # = density (cells/mm^3) * volume (mm^3)

@@ -218,12 ~~218,12 @@ if funcLocs and coordFunc == coord~~‘norm’: # if locations for this coordinate calculated using density function randLocs[:,icoord] = funcLocs

if sim.cfg.verbose and not funcLocs: print ‘Volume=%.4f, density=%.2f, numCells=%.0f’%(volume, self.tags[‘density’], self.tags[‘numCells’])
if sim.cfg.verbose and not funcLocs: print(‘Volume=%.4f, density=%.2f, numCells=%.0f’%(volume, self.tags[‘density’], self.tags[‘numCells’]))
for i in self._distributeCells(self.tags[‘numCells’])[sim.rank]: gid = sim.net.lastGid+i self.cellGids.append(gid) # add gid list of cells belonging to this population - not needed?
cellTags = {k: v for (k, v) in self.tags.iteritems() if k in sim.net.params.popTagsCopiedToCells} # copy all pop tags to cell tags, except those that are pop-specific
cellTags = {k: v for (k, v) in self.tags.items() if k in sim.net.params.popTagsCopiedToCells} # copy all pop tags to cell tags, except those that are pop-specific cellTags[‘pop’] = self.tags[‘pop’] cellTags[‘xnorm’] = randLocs[i,0] # calculate x location (um) cellTags[‘ynorm’] = randLocs[i,1] # calculate y location (um)

@@ -233,7 +233,7 @@ cellTags[‘z’] = sizeZ * randLocs[i,2] # calculate z location (um) cells.append(self.cellModelClass(gid, cellTags)) # instantiate Cell object if sim.cfg.verbose:

print(‘Cell %d/%d (gid=%d) of pop %s, pos=(%2.f, %2.f, %2.f), on node %d, ‘%(i, self.tags[‘numCells’]-1, gid, self.tags[‘pop’],cellTags[‘x’], cellTags[‘y’], cellTags[‘z’], sim.rank))
print((‘Cell %d/%d (gid=%d) of pop %s, pos=(%2.f, %2.f, %2.f), on node %d, ‘%(i, self.tags[‘numCells’]-1, gid, self.tags[‘pop’],cellTags[‘x’], cellTags[‘y’], cellTags[‘z’], sim.rank))) sim.net.lastGid = sim.net.lastGid + self.tags[‘numCells’] return cells

@@ -247,7 +247,7 @@

gid = sim.net.lastGid+i self.cellGids.append(gid) # add gid list of cells belonging to this population - not needed?

cellTags = {k: v for (k, v) in self.tags.iteritems() if k in sim.net.params.popTagsCopiedToCells} # copy all pop tags to cell tags, except those that are pop-specific
cellTags = {k: v for (k, v) in self.tags.items() if k in sim.net.params.popTagsCopiedToCells} # copy all pop tags to cell tags, except those that are pop-specific cellTags[‘pop’] = self.tags[‘pop’] cellTags.update(self.tags[‘cellsList’][i]) # add tags specific to this cells for coord in [‘x’,’y’,’z’]:

@@ -259,7 ~~259,7 @@ cellTags[coord~~‘norm’] = cellTags[coord] = 0 if ‘propList’ not in cellTags: cellTags[‘propList’] = [] # initalize list of property sets if doesn’t exist cells.append(self.cellModelClass(gid, cellTags)) # instantiate Cell object

if sim.cfg.verbose: print(‘Cell %d/%d (gid=%d) of pop %d, on node %d, ‘%(i, self.tags[‘numCells’]-1, gid, i, sim.rank))
if sim.cfg.verbose: print((‘Cell %d/%d (gid=%d) of pop %d, on node %d, ‘%(i, self.tags[‘numCells’]-1, gid, i, sim.rank))) sim.net.lastGid = sim.net.lastGid + len(self.tags[‘cellsList’]) return cells

@@ -290,7 +290,7 @@ for i in self._distributeCells(numCells)[sim.rank]: gid = sim.net.lastGid+i self.cellGids.append(gid) # add gid list of cells belonging to this population - not needed?

cellTags = {k: v for (k, v) in self.tags.iteritems() if k in sim.net.params.popTagsCopiedToCells} # copy all pop tags to cell tags, except those that are pop-specific
cellTags = {k: v for (k, v) in self.tags.items() if k in sim.net.params.popTagsCopiedToCells} # copy all pop tags to cell tags, except those that are pop-specific cellTags[‘pop’] = self.tags[‘pop’] cellTags[‘xnorm’] = gridLocs[i][0] / sim.net.params.sizeX # set x location (um) cellTags[‘ynorm’] = gridLocs[i][1] / sim.net.params.sizeY # set y location (um)

@@ -299,7 +299,7 @@ cellTags[‘y’] = gridLocs[i][1] # set y location (um) cellTags[‘z’] = gridLocs[i][2] # set z location (um) cells.append(self.cellModelClass(gid, cellTags)) # instantiate Cell object

if sim.cfg.verbose: print(‘Cell %d/%d (gid=%d) of pop %s, on node %d, ‘%(i, numCells, gid, self.tags[‘pop’], sim.rank))
if sim.cfg.verbose: print((‘Cell %d/%d (gid=%d) of pop %s, on node %d, ‘%(i, numCells, gid, self.tags[‘pop’], sim.rank))) sim.net.lastGid = sim.net.lastGid + numCells return cells

@@ -322,7 +322,7 @@ self.cellModelClass = sim.PointCell excludeTags = [‘pop’, ‘cellModel’, ‘cellType’, ‘numCells’, ‘density’, ‘xRange’, ‘yRange’, ‘zRange’, ‘xnormRange’, ‘ynormRange’, ‘znormRange’, ‘vref’]

params = {k: v for k,v in self.tags.iteritems() if k not in excludeTags}
params = {k: v for k,v in self.tags.items() if k not in excludeTags} self.tags[‘params’] = params for k in self.tags[‘params’]: self.tags.pop(k) sim.net.params.popTagsCopiedToCells.append(‘params’)

RefactoringTool: Writing converted netpyne/pop.py to netpyne_py3/pop.py. RefactoringTool: Refactored netpyne/sim.py — netpyne/sim.py (original) +++ netpyne/sim.py (refactored) @@ -14,13 +14,13 @@ netpyne.__gui__ = False

-from simFuncs import * -from neuromlFuncs import * -from wrappers import * -import analysis -from network import Network -from cell import CompartCell, PointCell, NML2Cell, NML2SpikeSource -from pop import Pop -import utils +from .simFuncs import * +from .neuromlFuncs import * +from .wrappers import * +from . import analysis +from .network import Network +from .cell import CompartCell, PointCell, NML2Cell, NML2SpikeSource +from .pop import Pop +from . import utils from neuron import h

RefactoringTool: Writing converted netpyne/sim.py to netpyne_py3/sim.py. RefactoringTool: Refactored netpyne/simFuncs.py — netpyne/simFuncs.py (original) +++ netpyne/simFuncs.py (refactored) @@ -17,14 +17,14 @@ import os from time import time from datetime import datetime -import cPickle as pk +import pickle as pk import hashlib from numbers import Number from copy import copy -from specs import Dict, ODict +from .specs import Dict, ODict from collections import OrderedDict from neuron import h, init # Import NEURON -import sim, specs +from . import sim, specs

@@ -120,7 +120,7 @@ else: return specs.NetParams(data[‘net’][‘params’]) else:

print(‘netParams not found in file %s’%(filename))
print((‘netParams not found in file %s’%(filename)))
pass

@@ -138,9 +138,9 @@ sim.net.allCells = data[‘net’][‘cells’] if instantiate:

cellsNode = [data[‘net’][‘cells’][i] for i in xrange(int(sim.rank), len(data[‘net’][‘cells’]), sim.nhosts)]
cellsNode = [data[‘net’][‘cells’][i] for i in range(int(sim.rank), len(data[‘net’][‘cells’]), sim.nhosts)] if sim.cfg.createPyStruct:
for popLoadLabel, popLoad in data[‘net’][‘pops’].iteritems():
for popLoadLabel, popLoad in data[‘net’][‘pops’].items(): pop = sim.Pop(popLoadLabel, popLoad[‘tags’]) pop.cellGids = popLoad[‘cellGids’] sim.net.pops[popLoadLabel] = pop

@@ -164,9 +164,9 @@ if sim.cfg.verbose: ’ Unable to load cell stims’

sim.net.cells.append(cell)

print(’ Created %d cells’ % (len(sim.net.cells)))
print(’ Created %d connections’ % (sum([len(c.conns) for c in sim.net.cells])))
print(’ Created %d stims’ % (sum([len(c.stims) for c in sim.net.cells])))
print((’ Created %d cells’ % (len(sim.net.cells))))
print((’ Created %d connections’ % (sum([len(c.conns) for c in sim.net.cells]))))
print((’ Created %d stims’ % (sum([len(c.stims) for c in sim.net.cells]))))
if sim.cfg.createNEURONObj:

@@ -185,13 +185,13 @@ except: if sim.cfg.verbose: ’ Unable to load instantiate cell conns or stims’

print(’ Added NEURON objects to %d cells’ % (len(sim.net.cells)))
print((’ Added NEURON objects to %d cells’ % (len(sim.net.cells))))
if sim.rank == 0 and sim.cfg.timing: sim.timing(‘stop’, ‘loadNetTime’)
print(’ Done; re-instantiate net time = %0.2f s’ % sim.timingData[‘loadNetTime’])
print((’ Done; re-instantiate net time = %0.2f s’ % sim.timingData[‘loadNetTime’])) else:
print(’ netCells and/or netPops not found in file %s’%(filename))
print((’ netCells and/or netPops not found in file %s’%(filename)))

############################################################################### @@ -206,7 +206,7 @@ else: return specs.SimConfig(data[‘simConfig’]) else:

print(’ simConfig not found in file %s’%(filename))
print((’ simConfig not found in file %s’%(filename))) pass

@@ -219,7 +219,7 @@ if ‘simData’ in data: sim.allSimData = data[‘simData’] else:

print(’ simData not found in file %s’%(filename))
print((’ simData not found in file %s’%(filename)))
pass

@@ -247,14 +247,14 @@

if ext == ‘pkl’: import pickle

print(‘Loading file %s … ’ % (filename))
print((‘Loading file %s … ’ % (filename))) with open(filename, ‘r’) as fileObj: data = pickle.load(fileObj)
elif ext == ‘dpk’: import gzip
print(‘Loading file %s … ’ % (filename))
print((‘Loading file %s … ’ % (filename))) #fn=sim.cfg.filename #.split(‘.’) #gzip.open(fn, ‘wb’).write(pk.dumps(dataSave)) # write compressed string print(‘NOT IMPLEMENTED!’)

@@ -262,14 +262,14 @@

elif ext == ‘json’: import json

print(‘Loading file %s … ’ % (filename))
print((‘Loading file %s … ’ % (filename))) with open(filename, ‘r’) as fileObj: data = json.load(fileObj, object_pairs_hook=OrderedDict)
elif ext == ‘mat’: from scipy.io import savemat
print(‘Loading file %s … ’ % (filename))
print((‘Loading file %s … ’ % (filename))) #savemat(sim.cfg.filename+’.mat’, replaceNoneObj(dataSave)) # replace None and {} with [] so can save in .mat format print(‘Finished saving!’)

@@ -277,14 +277,14 @@ elif ext == ‘saveHDF5’: #dataSaveUTF8 = _dict2utf8(replaceNoneObj(dataSave)) # replace None and {} with [], and convert to utf import hdf5storage

print(‘Loading file %s … ’ % (filename))
print((‘Loading file %s … ’ % (filename))) #hdf5storage.writes(dataSaveUTF8, filename=sim.cfg.filename+’.hdf5’) print(‘NOT IMPLEMENTED!’)
elif ext == ‘csv’: import csv
print(‘Loading file %s … ’ % (filename))
print((‘Loading file %s … ’ % (filename))) writer = csv.writer(open(sim.cfg.filename+’.csv’, ‘wb’)) #for dic in dataSave[‘simData’]:

@@ -293,7 +293,7 @@

elif ext == ‘dat’:

print(‘Loading file %s … ’ % (filename))
print((‘Loading file %s … ’ % (filename))) print(‘NOT IMPLEMENTED!’)

@@ -306,12 +306,12 @@

else:

print(‘Format not recognized for file %s’%(filename))
print((‘Format not recognized for file %s’%(filename))) return
if hasattr(sim, ‘rank’) and sim.rank == 0 and hasattr(sim, ‘cfg’) and sim.cfg.timing: sim.timing(‘stop’, ‘loadFileTime’)
print(’ Done; file loading time = %0.2f s’ % sim.timingData[‘loadFileTime’])
print((’ Done; file loading time = %0.2f s’ % sim.timingData[‘loadFileTime’]))

return data @@ -327,7 +327,7 @@

sim.clearObj([cell.__dict__ for cell in sim.net.cells]) sim.clearObj([stim for stim in sim.simData[‘stims’]])

for key in sim.simData.keys(): del sim.simData[key]
for key in list(sim.simData.keys()): del sim.simData[key] for c in sim.net.cells: del c for p in sim.net.pops: del p del sim.net.params

@@ -337,7 +337,7 @@ if sim.rank == 0: sim.clearObj([cell.__dict__ for cell in sim.net.allCells]) sim.clearObj([stim for stim in sim.allSimData[‘stims’]])

for key in sim.allSimData.keys(): del sim.allSimData[key]
for key in list(sim.allSimData.keys()): del sim.allSimData[key] for c in sim.net.allCells: del c for p in sim.net.allPops: del p del sim.net.allCells

@@ -380,7 +380,7 @@ elif type(obj) in [dict, Dict]: if objCopy == ‘ROOT’: objCopy = Dict()

for key,val in obj.iteritems():
for key,val in obj.items(): if type(val) in [list]: objCopy[key] = [] copyReplaceItemObj(val, keystart, newval, objCopy[key])

@@ -405,7 +405,7 @@ del item

elif type(obj) in [dict, Dict, ODict]:

for key in obj.keys():
for key in list(obj.keys()): val = obj[key] if type(val) in [list, dict, Dict, ODict]: clearObj(val)

@@ -422,7 +422,7 @@ replaceItemObj(item, keystart, newval)

elif type(obj) == dict:

for key,val in obj.iteritems():
for key,val in obj.items(): if type(val) in [list, dict]: replaceItemObj(val, keystart, newval) if key.startswith(keystart):

@@ -440,7 +440,7 @@ replaceFuncObj(item)

elif type(obj) == dict:

for key,val in obj.iteritems():
for key,val in obj.items(): if type(val) in [list, dict]: replaceFuncObj(val) if ‘func_name’ in dir(val): #hasattr(val,’func_name’): # avoid hasattr() since it creates key in Dicts()

@@ -458,7 +458,7 @@ replaceNoneObj(item)

elif type(obj) in [dict, Dict, ODict]:

for key,val in obj.iteritems():
for key,val in obj.items(): if type(val) in [list, dict, Dict, ODict]: replaceNoneObj(val) if val == None:

@@ -482,7 +482,7 @@ replaceDictODict(item)

elif type(obj) in [dict, OrderedDict, Dict, ODict]:

for key,val in obj.iteritems():
for key,val in obj.items(): if type(val) == Dict: obj[key] = val.todict() elif type(val) == ODict:

@@ -521,7 +521,7 @@ obj[obj.index(item)] = str(item)

elif type(obj) == dict or type(obj) == ODict:

for key,val in obj.iteritems():
for key,val in obj.items(): if type(val) in [list, dict, ODict]: tupleToStr(val) elif type(val) == tuple:

@@ -536,16 +536,16 @@ #unidict = {k.decode(‘utf8’): v.decode(‘utf8’) for k, v in strdict.items()} #print obj import collections

if isinstance(obj, basestring):
if isinstance(obj, str): return obj.decode(‘utf8’) elif isinstance(obj, collections.Mapping):
for key in obj.keys():
for key in list(obj.keys()): if isinstance(key, Number): obj[str(key).decode(‘utf8’)] = obj[key] obj.pop(key)
return dict(map(_dict2utf8, obj.iteritems()))
return dict(list(map(_dict2utf8, iter(obj.items())))) elif isinstance(obj, collections.Iterable):
return type(obj)(map(_dict2utf8, obj))
return type(obj)(list(map(_dict2utf8, obj))) else: return obj

@@ -565,7 +565,7 @@ import imp, __main__

if len(sys.argv) > 1:

print ‘\nReading command line arguments using syntax: python file.py [simConfig=filepath] [netParams=filepath]’
print(‘\nReading command line arguments using syntax: python file.py [simConfig=filepath] [netParams=filepath]’)
cfgPath = None netParamsPath = None

@@ -583,7 +583,7 @@ elif netParamsPath.endswith(‘py’): netParamsModule = imp.load_source(os.path.basename(netParamsPath).split(‘.’)[0], netParamsPath) netParams = netParamsModule.netParams

print ‘Importing netParams from %s’ %(netParamsPath)
print(‘Importing netParams from %s’ %(netParamsPath))
if not cfgPath: try:

@@ -591,7 +591,7 @@ cfg = cfgModule.cfg __main__.cfg = cfg except:

print ‘\nWarning: Could not load cfg from command line path or from default cfg.py’
print(‘\nWarning: Could not load cfg from command line path or from default cfg.py’) cfg = None
if not netParamsPath:

@@ -599,7 +599,7 @@ netParamsModule = imp.load_source(‘netParams’, ‘netParams.py’) netParams = netParamsModule.netParams except:

print ‘\nWarning: Could not load netParams from command line path or from default netParams.py’
print(‘\nWarning: Could not load netParams from command line path or from default netParams.py’) netParams = None
return cfg, netParams

@@ -618,7 +618,7 @@

if ‘plotRaster’ in sim.cfg.analysis: if isinstance(sim.cfg.analysis[‘plotRaster’],dict) and ‘include’ in sim.cfg.analysis[‘plotRaster’]:

netStimLabels = sim.net.params.stimSourceParams.keys()+[‘allNetStims’]
netStimLabels = list(sim.net.params.stimSourceParams.keys())+[‘allNetStims’] for item in sim.cfg.analysis[‘plotRaster’][‘include’]: if item in netStimLabels: sim.cfg.recordStim = True

@@ -629,7 +629,7 @@ sim.cfg.recordStim = True

elif (isinstance(sim.cfg.analysis[‘plotSpikeHist’],dict) and ‘include’ in sim.cfg.analysis[‘plotSpikeHist’]) :

netStimLabels = sim.net.params.stimSourceParams.keys()+[‘allNetStims’,’eachPop’]
netStimLabels = list(sim.net.params.stimSourceParams.keys())+[‘allNetStims’,’eachPop’] for item in sim.cfg.analysis[‘plotSpikeHist’][‘include’]: if item in netStimLabels: sim.cfg.recordStim = True

@@ -651,18 +651,18 @@

cellsRecord = getCellsList(sim.cfg.recordCells)+cellsPlot

for key in sim.cfg.recordTraces.keys(): sim.simData[key] = Dict() # create dict to store traces
for key in list(sim.cfg.recordTraces.keys()): sim.simData[key] = Dict() # create dict to store traces for cell in cellsRecord: cell.recordTraces() # call recordTraces function for each cell
cat = 0 total = 0 for key in sim.simData:
if sim.cfg.verbose: print(” Recording: %s:”%key)
if sim.cfg.verbose: print((” Recording: %s:”%key)) if len(sim.simData[key])>0: cat+=1 for k2 in sim.simData[key]:
if sim.cfg.verbose: print(” %s”%k2)
if sim.cfg.verbose: print((” %s”%k2)) total+=1
print(“Recording %s traces of %s types on node %i”%(total, cat, sim.rank))
print((“Recording %s traces of %s types on node %i”%(total, cat, sim.rank)))

timing(‘stop’, ‘setrecordTime’) @@ -689,11 +689,11 @@ elif isinstance(condition, int): # cell gid cellGids.append(condition)

elif isinstance(condition, basestring): # entire pop
elif isinstance(condition, str): # entire pop cellGids.extend(list(sim.net.pops[condition].cellGids))
elif isinstance(condition, tuple) or isinstance(condition, list): # subset of a pop with relative indices
cellsPop = [gid for gid,tags in allCellTags.iteritems() if tags[‘pop’]==condition[0]]
cellsPop = [gid for gid,tags in allCellTags.items() if tags[‘pop’]==condition[0]]
if isinstance(condition[1], list): cellGids.extend([gid for i,gid in enumerate(cellsPop) if i in condition[1]])

@@ -711,31 +711,31 @@ def setGlobals (): hParams = sim.cfg.hParams

cellGlobs = {k:v for k,v in hParams.iteritems()}
for cellRuleName, cellRule in sim.net.params.cellParams.iteritems():
for k,v in getattr(cellRule, ‘globals’, {}).iteritems():
cellGlobs = {k:v for k,v in hParams.items()}
for cellRuleName, cellRule in sim.net.params.cellParams.items():
for k,v in getattr(cellRule, ‘globals’, {}).items(): if k not in cellGlobs: cellGlobs[k] = v elif k in [‘celsius’, ‘v_init’, ‘clamp_resist’] and cellGlobs[k] != v: # exception if k == ‘v_init’: wrongVinit = [s[‘v_init’] for s in cellRule[‘secs’] if ‘v_init’ in s and s[‘v_init’] != v] # check if set inside secs if len(wrongVinit) > 0:
print ‘\nWarning: global variable %s=%s, but cellParams rule %s requires %s=%s’ % (k, str(cellGlobs[k]), cellRuleName, k, str(v))
print(‘\nWarning: global variable %s=%s, but cellParams rule %s requires %s=%s’ % (k, str(cellGlobs[k]), cellRuleName, k, str(v))) else:
print ‘\nWarning: global variable %s=%s, but cellParams rule %s requires %s=%s’ % (k, str(cellGlobs[k]), cellRuleName, k, str(v))
print(‘\nWarning: global variable %s=%s, but cellParams rule %s requires %s=%s’ % (k, str(cellGlobs[k]), cellRuleName, k, str(v))) elif k in cellGlobs and cellGlobs[k] != v:
print ‘\nError: global variable %s has different values (%s vs %s) in two cellParams rules’ % (k, str(v), str(cellGlobs[k]))
print(‘\nError: global variable %s has different values (%s vs %s) in two cellParams rules’ % (k, str(v), str(cellGlobs[k]))) sys.exit()
if sim.cfg.verbose and len(cellGlobs) > 0:
print ‘\nSetting h global variables …’
for key,val in cellGlobs.iteritems():
print(‘\nSetting h global variables …’)
for key,val in cellGlobs.items(): try: setattr(h, key, val) # set other h global vars (celsius, clamp_resist)
if sim.cfg.verbose: print(’ h.%s = %s’ % (key, str(val)))
if sim.cfg.verbose: print((’ h.%s = %s’ % (key, str(val)))) except:
print ‘\nError: could not set global %s = %s’ % (key, str(val))
print(‘\nError: could not set global %s = %s’ % (key, str(val)))

############################################################################### @@ -774,7 +774,7 @@

sim.pc.set_maxstep(10) mindelay = sim.pc.allreduce(sim.pc.set_maxstep(10), 2) # flag 2 returns minimum value

if sim.rank==0 and sim.cfg.verbose: print(‘Minimum delay (time-step for queue exchange) is %.2f’%(mindelay))
if sim.rank==0 and sim.cfg.verbose: print((‘Minimum delay (time-step for queue exchange) is %.2f’%(mindelay))) sim.pc.setup_transfer() # setup transfer of source_var to target_var

@@ -782,7 +782,7 @@ def printRunTime(): h(‘objref cvode’) h(‘cvode = new CVode()’)

for i in xrange(int(sim.cfg.printRunTime*1000.0), int(sim.cfg.duration), int(sim.cfg.printRunTime*1000.0)):
for i in range(int(sim.cfg.printRunTime*1000.0), int(sim.cfg.duration), int(sim.cfg.printRunTime*1000.0)): h.cvode.event(i, ‘print ’ + str(i/1000.0) + ‘,”s”’)
sim.printRunTime = printRunTime

@@ -812,14 +812,14 @@ preRun() init()

if sim.rank == 0: print(‘\nRunning simulation for %s ms…’%sim.cfg.duration)
if sim.rank == 0: print((‘\nRunning simulation for %s ms…’%sim.cfg.duration)) sim.pc.psolve(sim.cfg.duration)
sim.pc.barrier() # Wait for all hosts to get to this point timing(‘stop’, ‘runTime’) if sim.rank==0:
print(’ Done; run time = %0.2f s; real-time ratio: %0.2f.’ %
(sim.timingData[‘runTime’], sim.cfg.duration/1000/sim.timingData[‘runTime’]))
print((’ Done; run time = %0.2f s; real-time ratio: %0.2f.’ %
(sim.timingData[‘runTime’], sim.cfg.duration/1000/sim.timingData[‘runTime’])))

############################################################################### @@ -839,8 +839,8 @@ sim.pc.barrier() # Wait for all hosts to get to this point timing(‘stop’, ‘runTime’) if sim.rank==0:

print(’ Done; run time = %0.2f s; real-time ratio: %0.2f.’ %
(sim.timingData[‘runTime’], sim.cfg.duration/1000/sim.timingData[‘runTime’]))
print((’ Done; run time = %0.2f s; real-time ratio: %0.2f.’ %
(sim.timingData[‘runTime’], sim.cfg.duration/1000/sim.timingData[‘runTime’])))

############################################################################### @@ -887,35 +887,35 @@ for cell in sim.net.cells: cell.conns = []

simDataVecs = [‘spkt’,’spkid’,’stims’]+sim.cfg.recordTraces.keys()
simDataVecs = [‘spkt’,’spkid’,’stims’]+list(sim.cfg.recordTraces.keys()) if sim.nhosts > 1: # only gather if >1 nodes
netPopsCellGids = {popLabel: list(pop.cellGids) for popLabel,pop in sim.net.pops.iteritems()}
netPopsCellGids = {popLabel: list(pop.cellGids) for popLabel,pop in sim.net.pops.items()}
if getattr(sim.cfg, ‘gatherOnlySimData’, False): nodeData = {‘simData’: sim.simData} data = [None]*sim.nhosts data[0] = {}
for k,v in nodeData.iteritems():
for k,v in nodeData.items(): data[0][k] = v gather = sim.pc.py_alltoall(data) sim.pc.barrier()
if sim.rank == 0: # simData
print ’ Gathering only sim data…’
print(’ Gathering only sim data…’) sim.allSimData = Dict()
for k in gather[0][‘simData’].keys(): # initialize all keys of allSimData dict
for k in list(gather[0][‘simData’].keys()): # initialize all keys of allSimData dict sim.allSimData[k] = {}
for node in gather: # concatenate data from each node
for key,val in node[‘simData’].iteritems(): # update simData dics of dics of h.Vector
for key,val in node[‘simData’].items(): # update simData dics of dics of h.Vector if key in simDataVecs: # simData dicts that contain Vectors if isinstance(val,dict):
for cell,val2 in val.iteritems():
for cell,val2 in val.items(): if isinstance(val2,dict): sim.allSimData[key].update(Dict({cell:Dict()}))
for stim,val3 in val2.iteritems():
for stim,val3 in val2.items(): sim.allSimData[key][cell].update({stim:list(val3)}) # udpate simData dicts which are dicts of dicts of Vectors (eg. [‘stim’][‘cell_1’][‘backgrounsd’]=h.Vector) else: sim.allSimData[key].update({cell:list(val2)}) # udpate simData dicts which are dicts of Vectors (eg. [‘v’][‘cell_1’]=h.Vector)

@@ -926,7 +926,7 @@

sim.net.allPops = ODict() # pops

for popLabel,pop in sim.net.pops.iteritems(): sim.net.allPops[popLabel] = pop.__getstate__() # can’t use dict comprehension for OrderedDict
for popLabel,pop in sim.net.pops.items(): sim.net.allPops[popLabel] = pop.__getstate__() # can’t use dict comprehension for OrderedDict
sim.net.allCells = [c.__dict__ for c in sim.net.cells]

@@ -935,33 +935,33 @@ nodeData = {‘netCells’: [c.__getstate__() for c in sim.net.cells], ‘netPopsCellGids’: netPopsCellGids, ‘simData’: sim.simData} data = [None]*sim.nhosts data[0] = {}

for k,v in nodeData.iteritems():
for k,v in nodeData.items(): data[0][k] = v gather = sim.pc.py_alltoall(data) sim.pc.barrier() if sim.rank == 0: allCells = [] allPops = ODict()
for popLabel,pop in sim.net.pops.iteritems(): allPops[popLabel] = pop.__getstate__() # can’t use dict comprehension for OrderedDict
for popLabel,pop in sim.net.pops.items(): allPops[popLabel] = pop.__getstate__() # can’t use dict comprehension for OrderedDict allPopsCellGids = {popLabel: [] for popLabel in netPopsCellGids} sim.allSimData = Dict()
for k in gather[0][‘simData’].keys(): # initialize all keys of allSimData dict
for k in list(gather[0][‘simData’].keys()): # initialize all keys of allSimData dict sim.allSimData[k] = {}
for node in gather: # concatenate data from each node allCells.extend(node[‘netCells’]) # extend allCells list
for popLabel,popCellGids in node[‘netPopsCellGids’].iteritems():
for popLabel,popCellGids in node[‘netPopsCellGids’].items(): allPopsCellGids[popLabel].extend(popCellGids)
for key,val in node[‘simData’].iteritems(): # update simData dics of dics of h.Vector
for key,val in node[‘simData’].items(): # update simData dics of dics of h.Vector if key in simDataVecs: # simData dicts that contain Vectors if isinstance(val,dict):
for cell,val2 in val.iteritems():
for cell,val2 in val.items(): if isinstance(val2,dict): sim.allSimData[key].update(Dict({cell:Dict()}))
for stim,val3 in val2.iteritems():
for stim,val3 in val2.items(): sim.allSimData[key][cell].update({stim:list(val3)}) # udpate simData dicts which are dicts of dicts of Vectors (eg. [‘stim’][‘cell_1’][‘backgrounsd’]=h.Vector) else: sim.allSimData[key].update({cell:list(val2)}) # udpate simData dicts which are dicts of Vectors (eg. [‘v’][‘cell_1’]=h.Vector)

@@ -972,7 +972,7 @@

sim.net.allCells = sorted(allCells, key=lambda k: k[‘gid’])

for popLabel,pop in allPops.iteritems():
for popLabel,pop in allPops.items(): pop[‘cellGids’] = sorted(allPopsCellGids[popLabel]) sim.net.allPops = allPops

@@ -993,17 +993,17 @@ else: sim.net.allCells = [c.__dict__ for c in sim.net.cells] sim.net.allPops = ODict()

for popLabel,pop in sim.net.pops.iteritems(): sim.net.allPops[popLabel] = pop.__getstate__() # can’t use dict comprehension for OrderedDict
for popLabel,pop in sim.net.pops.items(): sim.net.allPops[popLabel] = pop.__getstate__() # can’t use dict comprehension for OrderedDict sim.allSimData = Dict()
for k in sim.simData.keys(): # initialize all keys of allSimData dict
for k in list(sim.simData.keys()): # initialize all keys of allSimData dict sim.allSimData[k] = Dict()
for key,val in sim.simData.iteritems(): # update simData dics of dics of h.Vector
for key,val in sim.simData.items(): # update simData dics of dics of h.Vector if key in simDataVecs: # simData dicts that contain Vectors if isinstance(val,dict):
for cell,val2 in val.iteritems():
for cell,val2 in val.items(): if isinstance(val2,dict): sim.allSimData[key].update(Dict({cell:Dict()}))
for stim,val3 in val2.iteritems():
for stim,val3 in val2.items(): sim.allSimData[key][cell].update({stim:list(val3)}) # udpate simData dicts which are dicts of dicts of Vectors (eg. [‘stim’][‘cell_1’][‘backgrounsd’]=h.Vector) else: sim.allSimData[key].update({cell:list(val2)}) # udpate simData dicts which are dicts of Vectors (eg. [‘v’][‘cell_1’]=h.Vector)

@@ -1016,7 +1016,7 @@ sim.pc.barrier() if sim.rank == 0: timing(‘stop’, ‘gatherTime’)

if sim.cfg.timing: print(’ Done; gather time = %0.2f s.’ % sim.timingData[‘gatherTime’])
if sim.cfg.timing: print((’ Done; gather time = %0.2f s.’ % sim.timingData[‘gatherTime’]))
print(‘\nAnalyzing…’) sim.totalSpikes = len(sim.allSimData[‘spkt’])

@@ -1035,16 +1035,16 @@ sim.connsPerCell = 0 sim.synsPerCell = 0

print(’ Cells: %i’ % (sim.numCells) )
print(’ Connections: %i (%0.2f per cell)’ % (sim.totalConnections, sim.connsPerCell))
print((’ Cells: %i’ % (sim.numCells) ))
print((’ Connections: %i (%0.2f per cell)’ % (sim.totalConnections, sim.connsPerCell))) if sim.totalSynapses != sim.totalConnections:
print(’ Synaptic contacts: %i (%0.2f per cell)’ % (sim.totalSynapses, sim.synsPerCell))
print((’ Synaptic contacts: %i (%0.2f per cell)’ % (sim.totalSynapses, sim.synsPerCell))) if ‘runTime’ in sim.timingData:
print(’ Spikes: %i (%0.2f Hz)’ % (sim.totalSpikes, sim.firingRate))
print((’ Spikes: %i (%0.2f Hz)’ % (sim.totalSpikes, sim.firingRate))) if sim.cfg.printPopAvgRates: sim.allSimData[‘popRates’] = sim.popAvgRates()
print(’ Simulated time: %0.1f s; %i workers’ % (sim.cfg.duration/1e3, sim.nhosts))
print(’ Run time: %0.2f s’ % (sim.timingData[‘runTime’]))
print((’ Simulated time: %0.1f s; %i workers’ % (sim.cfg.duration/1e3, sim.nhosts)))
print((’ Run time: %0.2f s’ % (sim.timingData[‘runTime’])))
sim.allSimData[‘avgRate’] = sim.firingRate # save firing rate

@@ -1056,7 +1056,7 @@ ############################################################################### def popAvgRates (trange = None, show = True): if not hasattr(sim, ‘allSimData’) or ‘spkt’ not in sim.allSimData:

print ‘Error: sim.allSimData not available; please call sim.gatherData()’
print(‘Error: sim.allSimData not available; please call sim.gatherData()’) return None
spkts = sim.allSimData[‘spkt’]

@@ -1065,7 +1065,7 @@ if not trange: trange = [0, sim.cfg.duration] else:

spkids,spkts = zip(*[(spkid,spkt) for spkid,spkt in zip(spkids,spkts) if trange[0] <= spkt <= trange[1]])
spkids,spkts = list(zip(*[(spkid,spkt) for spkid,spkt in zip(spkids,spkts) if trange[0] <= spkt <= trange[1]]))
avgRates = Dict() for pop in sim.net.allPops:

@@ -1073,7 +1073,7 @@ if numCells > 0: tsecs = float((trange[1]-trange[0]))/1000.0 avgRates[pop] = len([spkid for spkid in spkids if sim.net.allCells[int(spkid)][‘tags’][‘pop’]==pop])/numCells/tsecs

print ’ %s : %.3f Hz’%(pop, avgRates[pop])
print(’ %s : %.3f Hz’%(pop, avgRates[pop]))
return avgRates

@@ -1088,13 +1088,13 @@ avg_comp_time = sim.pc.allreduce(computation_time, 1)/sim.nhosts load_balance = avg_comp_time/max_comp_time

print ‘node:’,sim.rank,’ comp_time:’,computation_time
print(‘node:’,sim.rank,’ comp_time:’,computation_time) if sim.rank==0:
print ‘max_comp_time:’, max_comp_time
print ‘min_comp_time:’, min_comp_time
print ‘avg_comp_time:’, avg_comp_time
print ‘load_balance:’,load_balance
print ‘\nspike exchange time (run_time-comp_time): ‘, sim.timingData[‘runTime’] - max_comp_time
print(‘max_comp_time:’, max_comp_time)
print(‘min_comp_time:’, min_comp_time)
print(‘avg_comp_time:’, avg_comp_time)
print(‘load_balance:’,load_balance)
print(‘\nspike exchange time (run_time-comp_time): ‘, sim.timingData[‘runTime’] - max_comp_time)
return [max_comp_time, min_comp_time, avg_comp_time, load_balance]

@@ -1111,7 +1111,7 @@ nodeData = {‘netCells’: [c.__getstate__() for c in sim.net.cells]} data = [None]*sim.nhosts data[0] = {}

for k,v in nodeData.iteritems():
for k,v in nodeData.items(): data[0][k] = v gather = sim.pc.py_alltoall(data) sim.pc.barrier()

@@ -1153,7 +1153,7 @@

if isinstance(sim.cfg.backupCfgFile, list) and len(sim.cfg.backupCfgFile) == 2: simName = sim.cfg.simLabel if sim.cfg.simLabel else os.path.basename(sim.cfg.filename)

print(‘Copying cfg file %s … ’ % simName)
print((‘Copying cfg file %s … ’ % simName)) source = sim.cfg.backupCfgFile[0] targetFolder = sim.cfg.backupCfgFile[1]

@@ -1161,11 +1161,11 @@ os.mkdir(targetFolder) except OSError: if not os.path.exists(targetFolder):

print ’ Could not create target folder: %s’ % (targetFolder)
print(’ Could not create target folder: %s’ % (targetFolder))
targetFile = targetFolder + ‘/’ + simName + ‘_cfg.py’ if os.path.exists(targetFile):
print ’ Removing prior cfg file’ , targetFile
print(’ Removing prior cfg file’ , targetFile) os.system(‘rm ’ + targetFile) os.system(‘cp ’ + source + ’ ’ + targetFile)

@@ -1176,7 +1176,7 @@ try: os.mkdir(targetFolder) except OSError:

print ’ Could not create target folder: %s’ % (targetFolder)
print(’ Could not create target folder: %s’ % (targetFolder))
if not include: include = sim.cfg.saveDataInclude

@@ -1204,7 +1204,7 @@ if sim.cfg.savePickle: import pickle dataSave = replaceDictODict(dataSave)

print(‘Saving output as %s … ’ % (sim.cfg.filename+’.pkl’))
print((‘Saving output as %s … ’ % (sim.cfg.filename+’.pkl’))) with open(sim.cfg.filename+’.pkl’, ‘wb’) as fileObj: pickle.dump(dataSave, fileObj) print(‘Finished saving!’)

@@ -1212,7 +1212,7 @@

if sim.cfg.saveDpk: import gzip

print(‘Saving output as %s … ’ % (sim.cfg.filename+’.dpk’))
print((‘Saving output as %s … ’ % (sim.cfg.filename+’.dpk’))) fn=sim.cfg.filename #.split(‘.’) gzip.open(fn, ‘wb’).write(pk.dumps(dataSave)) # write compressed string print(‘Finished saving!’)

@@ -1221,7 +1221,7 @@ if sim.cfg.saveJson: import json #dataSave = replaceDictODict(dataSave) # not required since json saves as dict

print(‘Saving output as %s … ’ % (sim.cfg.filename+’.json ‘))
print((‘Saving output as %s … ’ % (sim.cfg.filename+’.json ‘))) with open(sim.cfg.filename+’.json’, ‘w’) as fileObj: json.dump(dataSave, fileObj) print(‘Finished saving!’)

@@ -1229,7 +1229,7 @@

if sim.cfg.saveMat: from scipy.io import savemat

print(‘Saving output as %s … ’ % (sim.cfg.filename+’.mat’))
print((‘Saving output as %s … ’ % (sim.cfg.filename+’.mat’))) savemat(sim.cfg.filename+’.mat’, tupleToStr(replaceNoneObj(dataSave))) # replace None and {} with [] so can save in .mat format print(‘Finished saving!’)

@@ -1237,7 +1237,7 @@ if sim.cfg.saveHDF5: dataSaveUTF8 = _dict2utf8(replaceNoneObj(dataSave)) # replace None and {} with [], and convert to utf import hdf5storage

print(‘Saving output as %s… ’ % (sim.cfg.filename+’.hdf5’))
print((‘Saving output as %s… ’ % (sim.cfg.filename+’.hdf5’))) hdf5storage.writes(dataSaveUTF8, filename=sim.cfg.filename+’.hdf5’) print(‘Finished saving!’)

@@ -1245,7 +1245,7 @@ if sim.cfg.saveCSV: if ‘simData’ in dataSave: import csv

print(‘Saving output as %s … ’ % (sim.cfg.filename+’.csv’))
print((‘Saving output as %s … ’ % (sim.cfg.filename+’.csv’))) writer = csv.writer(open(sim.cfg.filename+’.csv’, ‘wb’)) for dic in dataSave[‘simData’]: for values in dic:

@@ -1255,12 +1255,12 @@

if sim.cfg.saveDat: traces = sim.cfg.recordTraces

for ref in traces.keys():
for cellid in sim.allSimData[ref].keys():
for ref in list(traces.keys()):
for cellid in list(sim.allSimData[ref].keys()): dat_file_name = ‘%s_%s.dat’%(ref,cellid) dat_file = open(dat_file_name, ‘w’) trace = sim.allSimData[ref][cellid]
print(“Saving %i points of data on: %s:%s to %s”%(len(trace),ref,cellid,dat_file_name))
print((“Saving %i points of data on: %s:%s to %s”%(len(trace),ref,cellid,dat_file_name))) for i in range(len(trace)): dat_file.write(‘%s\t%s\n’%((i*sim.cfg.dt/1000),trace[i]/1000))

@@ -1269,14 +1269,14 @@

if sim.cfg.timing: timing(‘stop’, ‘saveTime’)

print(’ Done; saving time = %0.2f s.’ % sim.timingData[‘saveTime’])
print((’ Done; saving time = %0.2f s.’ % sim.timingData[‘saveTime’])) if sim.cfg.timing and sim.cfg.saveTiming: import pickle with open(‘timing.pkl’, ‘wb’) as file: pickle.dump(sim.timing, file)

for key in dataSave.keys():
for key in list(dataSave.keys()): del dataSave[key] del dataSave

@@ -1326,15 +1326,15 @@

if sim.rank == 0: # and checkMemory: import resource

print ‘\nMEMORY -----------------------’
print ‘Sections: ’
print h.topology()
print ‘NetCons: ’
print len(h.List(“NetCon”))
print ‘NetStims:’
print len(h.List(“NetStim”))
print ‘\n Memory usage: %s \n’ % resource.getrusage(resource.RUSAGE_SELF).ru_maxrss
print(‘\nMEMORY -----------------------‘)
print(‘Sections: ‘)
print(h.topology())
print(‘NetCons: ‘)
print(len(h.List(“NetCon”)))
print(‘NetStims:’)
print(len(h.List(“NetStim”)))
print(‘\n Memory usage: %s \n’ % resource.getrusage(resource.RUSAGE_SELF).ru_maxrss)
print ‘--------------------------------\n’

-

print(‘--------------------------------\n’)

+ RefactoringTool: Writing converted netpyne/simFuncs.py to netpyne_py3/simFuncs.py. RefactoringTool: Refactored netpyne/specs.py — netpyne/specs.py (original) +++ netpyne/specs.py (refactored) @@ -66,18 +66,18 @@

def fromdict(self, d): d = self.dotify(d)

for k,v in d.iteritems():
for k,v in d.items(): self[k] = v
def __repr__(self):
keys = self.keys()
keys = list(self.keys()) args = ‘, ‘.join([‘%s: %r’ % (key, self[key]) for key in keys]) return ‘{%s}’ % (args)

def dotify(self, x): if isinstance(x, dict):

return Dict( (k, self.dotify(v)) for k,v in x.iteritems() )
return Dict( (k, self.dotify(v)) for k,v in x.items() ) elif isinstance(x, (list, tuple)): return type(x)( self.dotify(v) for v in x ) else:

@@ -85,7 +85,7 @@

def undotify(self, x): if isinstance(x, dict):

return dict( (k, self.undotify(v)) for k,v in x.iteritems() )
return dict( (k, self.undotify(v)) for k,v in x.items() ) elif isinstance(x, (list, tuple)): return type(x)( self.undotify(v) for v in x ) else:

@@ -166,20 +166,20 @@

def fromOrderedDict(self, d): d = self.dotify(d)

for k,v in d.iteritems():
for k,v in d.items(): self[k] = v
def __repr__(self):
keys = self.keys()
keys = list(self.keys()) args = ‘, ‘.join([‘%s: %r’ % (key, self[key]) for key in keys]) return ‘{%s}’ % (args)

def dotify(self, x): if isinstance(x, OrderedDict):

return ODict( (k, self.dotify(v)) for k,v in x.iteritems() )
return ODict( (k, self.dotify(v)) for k,v in x.items() ) elif isinstance(x, dict):
return Dict( (k, self.dotify(v)) for k,v in x.iteritems() )
return Dict( (k, self.dotify(v)) for k,v in x.items() ) elif isinstance(x, (list, tuple)): return type(x)( self.dotify(v) for v in x ) else:

@@ -187,9 +187,9 @@

def undotify(self, x): if isinstance(x, OrderedDict):

return OrderedDict( (k, self.undotify(v)) for k,v in x.iteritems() )
return OrderedDict( (k, self.undotify(v)) for k,v in x.items() ) elif isinstance(x, dict):
return dict( (k, self.undotify(v)) for k,v in x.iteritems() )
return dict( (k, self.undotify(v)) for k,v in x.items() ) elif isinstance(x, (list, tuple)): return type(x)( self.undotify(v) for v in x ) else:

@@ -249,7 +249,7 @@

if netParamsDict:

for k,v in netParamsDict.iteritems():
for k,v in netParamsDict.items(): if isinstance(v, OrderedDict): setattr(self, k, ODict(v)) elif isinstance(v, dict):

@@ -268,14 +268,14 @@ os.mkdir(folder) except OSError: if not os.path.exists(folder):

print ’ Could not create’, folder
print(’ Could not create’, folder)
dataSave = {‘netParams’: self.__dict__}

if ext == ‘json’: import json
print(‘Saving netParams to %s … ’ % (filename))
print((‘Saving netParams to %s … ’ % (filename))) with open(filename, ‘w’) as fileObj: json.dump(dataSave, fileObj, indent=4, sort_keys=True)

@@ -334,12 +334,12 @@ if somaAtOrigin: somaSec = next((sec for sec in cellRule[‘secs’] if ‘soma’ in sec), None) if not somaSec or not ‘pt3d’ in cellRule[‘secs’][somaSec][‘geom’]:

print ‘Warning: cannot place soma at origin because soma does not exist or does not contain pt3d’
print(‘Warning: cannot place soma at origin because soma does not exist or does not contain pt3d’) return soma3d = cellRule[‘secs’][somaSec][‘geom’][‘pt3d’] midpoint = int(len(soma3d)/2) somaX, somaY, somaZ = soma3d[midpoint][0:3]
for sec in cellRule[‘secs’].values():
for sec in list(cellRule[‘secs’].values()): for i,pt3d in enumerate(sec[‘geom’][‘pt3d’]): sec[‘geom’][‘pt3d’][i] = (pt3d[0] - somaX, pt3d[1] - somaY, pt3d[2] - somaZ, pt3d[3])

@@ -362,15 +362,15 @@ if label in self.cellParams: cellRule = self.cellParams[label] else:

print ‘Error adding secList: netParams.cellParams does not contain %s’ % (label)
print(‘Error adding secList: netParams.cellParams does not contain %s’ % (label)) return
if not isinstance(somaDist, list) or len(somaDist) != 2:
print ‘Error adding secList: somaDist should be a list with 2 elements’
print(‘Error adding secList: somaDist should be a list with 2 elements’) return
secList = []
for secName, sec in cellRule.secs.iteritems():
for secName, sec in cellRule.secs.items(): if ‘pt3d’ in sec[‘geom’]: pt3d = sec[‘geom’][‘pt3d’] midpoint = int(len(pt3d)/2)

@@ -380,7 +380,7 @@ secList.append(secName)

else:

print ‘Error adding secList: Sections do not contain 3d points’
print(‘Error adding secList: Sections do not contain 3d points’) return
cellRule.secLists[secListName] = list(secList)

@@ -390,15 +390,15 @@ if label in self.cellParams: cellRule = self.cellParams[label] else:

print ‘Error renaming section: netParams.cellParams does not contain %s’ % (label)
print(‘Error renaming section: netParams.cellParams does not contain %s’ % (label)) return
if oldSec not in cellRule[‘secs’]:
print ‘Error renaming section: cellRule does not contain section %s’ % (label)
print(‘Error renaming section: cellRule does not contain section %s’ % (label)) return
cellRule[‘secs’][newSec] = cellRule[‘secs’].pop(oldSec) # replace sec name
for sec in cellRule[‘secs’].values(): # replace appearences in topol
for sec in list(cellRule[‘secs’].values()): # replace appearences in topol if sec[‘topol’].get(‘parentSec’) == oldSec: sec[‘topol’][‘parentSec’] = newSec

@@ -407,19 +407,19 @@ if label in self.cellParams: cellRule = self.cellParams[label] else:

print ‘Error adding weightNorm: netParams.cellParams does not contain %s’ % (label)
print(‘Error adding weightNorm: netParams.cellParams does not contain %s’ % (label)) return
with open(fileName, ‘r’) as fileObj: weightNorm = pickle.load(fileObj)

try:
somaSec = next((k for k in weightNorm.keys() if k.startswith(‘soma’)),None)
somaSec = next((k for k in list(weightNorm.keys()) if k.startswith(‘soma’)),None) somaWeightNorm = weightNorm[somaSec][0] except:
print ‘Error setting weightNorm: no soma section available to set threshold’
return
for sec, wnorm in weightNorm.iteritems():
print(‘Error setting weightNorm: no soma section available to set threshold’)
return
for sec, wnorm in weightNorm.items(): if sec in cellRule[‘secs’]: wnorm = [min(wn,threshold*somaWeightNorm) for wn in wnorm] cellRule[‘secs’][sec][‘weightNorm’] = wnorm # add weight normalization factors for each section

@@ -430,7 +430,7 @@ if label in self.cellParams: cellRule = self.cellParams[label] else:

print ‘Error saving: netParams.cellParams does not contain %s’ % (label)
print(‘Error saving: netParams.cellParams does not contain %s’ % (label)) return with open(fileName, ‘w’) as fileObj: pickle.dump(cellRule, fileObj)

@@ -445,7 +445,7 @@

def todict(self):

from sim import replaceDictODict
from .sim import replaceDictODict return replaceDictODict(self.__dict__)

@@ -504,7 +504,7 @@

if simConfigDict:

for k,v in simConfigDict.iteritems():
for k,v in simConfigDict.items(): if isinstance(v, OrderedDict): setattr(self, k, ODict(v)) elif isinstance(v, dict):

@@ -523,14 +523,14 @@ os.mkdir(folder) except OSError: if not os.path.exists(folder):

print ’ Could not create’, folder
print(’ Could not create’, folder)
dataSave = {‘simConfig’: self.__dict__}

if ext == ‘json’: import json
print(‘Saving simConfig to %s … ’ % (filename))
print((‘Saving simConfig to %s … ’ % (filename))) with open(filename, ‘w’) as fileObj: json.dump(dataSave, fileObj, indent=4, sort_keys=True)

@@ -538,6 +538,6 @@ self.analysis[func] = params

def todict(self):

from sim import replaceDictODict
from .sim import replaceDictODict return replaceDictODict(self.__dict__)

RefactoringTool: Writing converted netpyne/specs.py to netpyne_py3/specs.py. RefactoringTool: Refactored netpyne/utils.py — netpyne/utils.py (original) +++ netpyne/utils.py (refactored) @@ -39,18 +39,18 @@ if filePath not in sys.path: # add to path if not there (need to import module) sys.path.insert(0, filePath) moduleName = fileNameOnly.split(‘.py’)[0] # remove .py to obtain module name

exec(‘import ‘+ moduleName + ’ as tempModule’) in locals() # import module dynamically
exec((‘import ‘+ moduleName + ’ as tempModule’), locals()) # import module dynamically modulePointer = tempModule paramLabels = getattr(modulePointer, labels) # tuple with labels paramValues = getattr(modulePointer, values) # variable with paramValues if key: # if paramValues = dict paramValues = paramValues[key]
params = dict(zip(paramLabels, paramValues))
params = dict(list(zip(paramLabels, paramValues))) sys.path.remove(filePath) except:
print “Error loading cell parameter values from ” + fileName
else:
print “Trying to import izhi params from a file without the .py extension”
print(“Error loading cell parameter values from ” + fileName)
else:
print(“Trying to import izhi params from a file without the .py extension”) return params

@@ -60,13 +60,13 @@ for i, mechtype in enumerate([‘mechs’,’pointps’]): mt = h.MechanismType(i) # either distributed mechs (0) or point process (1) varList[mechtype] = {}

for j in xrange(int(mt.count())):
for j in range(int(mt.count())): mt.select(j) mt.selected(msname) ms = h.MechanismStandard(msname[0], 1) # PARAMETER (modifiable) varList[mechtype][msname[0]] = [] propName = h.ref(”)
for var in xrange(int(ms.count())):
for var in range(int(ms.count())): k = ms.name(propName, var) varList[mechtype][msname[0]].append(propName[0]) return varList

@@ -104,7 +104,7 @@

def setGlobals (glob):

for k,v in glob.iteritems():
for k,v in glob.items(): setattr(h, k, v)

@@ -118,10 +118,10 @@

def _equal_dicts (d1, d2, ignore_keys): ignored = set(ignore_keys)

for k1, v1 in d1.iteritems():
for k1, v1 in d1.items(): if k1 not in ignored and (k1 not in d2 or d2[k1] != v1): return False
for k2, v2 in d2.iteritems():
for k2, v2 in d2.items(): if k2 not in ignored and k2 not in d1: return False return True

@@ -135,7 +135,7 @@ except KeyError: pass

for mod in modules.values():
for mod in list(modules.values()): try: delattr(mod, modname) except:

@@ -144,7 +144,7 @@ def importCell (fileName, cellName, cellArgs = None, cellInstance = False): h.initnrn() varList = mechVarList() # list of properties for all density mechanisms and point processes

origGlob = getGlobals(varList[‘mechs’].keys()+varList[‘pointps’].keys())
origGlob = getGlobals(list(varList[‘mechs’].keys())+list(varList[‘pointps’].keys()))
if cellArgs is None: cellArgs = [] # Define as empty list if not otherwise defined

@@ -166,7 +166,7 @@ if filePath not in sys.path: # add to path if not there (need to import module) sys.path.insert(0, filePath) moduleName = fileNameOnly.split(‘.py’)[0] # remove .py to obtain module name

exec(‘import ’ + moduleName + ’ as tempModule’) in globals(), locals() # import module dynamically
exec((‘import ’ + moduleName + ’ as tempModule’), globals(), locals()) # import module dynamically modulePointer = tempModule if isinstance(cellArgs, dict): cell = getattr(modulePointer, cellName)(**cellArgs) # create cell using template, passing dict with args

@@ -174,7 +174,7 @@ cell = getattr(modulePointer, cellName)(*cellArgs) # create cell using template, passing list with args sys.path.remove(filePath) else:

print “File name should be either .hoc or .py file”
print(“File name should be either .hoc or .py file”) return
secDic, secListDic, synMechs, globs = getCellParams(cell, varList, origGlob)

@@ -193,7 +193,7 @@

”’ Import cell from HOC template or python file into framework format (dict of sections, with geom, topol, mechs, syns)”’ if fileName.endswith(‘.hoc’) or fileName.endswith(‘.tem’):

print ‘Importing from .hoc network not yet supported’
print(‘Importing from .hoc network not yet supported’) return

@@ -207,18 +207,18 @@ sys.path.insert(0, filePath) moduleName = fileNameOnly.split(‘.py’)[0] # remove .py to obtain module name os.chdir(filePath)

print ‘\nRunning network in %s to import cells into NetPyNE …\n’%(fileName)
print(‘\nRunning network in %s to import cells into NetPyNE …\n’%(fileName)) from neuron import load_mechanisms load_mechanisms(filePath)
exec(‘import ’ + moduleName + ’ as tempModule’) in globals(), locals() # import module dynamically
exec((‘import ’ + moduleName + ’ as tempModule’), globals(), locals()) # import module dynamically modulePointer = tempModule sys.path.remove(filePath) else:
print “File name should be either .hoc or .py file”
print(“File name should be either .hoc or .py file”) return
for label, conds, cellName in zip(labelList, condsList, cellNamesList):
print ‘\nImporting %s from %s …’%(cellName, fileName)
print(‘\nImporting %s from %s …’%(cellName, fileName)) exec(‘cell = tempModule’ + ‘.’ + cellName) #cell = getattr(modulePointer, cellName) # get cell object secs, secLists, synMechs = getCellParams(cell)

@@ -253,7 +253,7 @@

dirCellSecNames = {} for sec in secs:

dirCellSecNames.update({hname: name for hname,name in dirCellHnames.iteritems() if hname == sec.hname()})
dirCellSecNames.update({hname: name for hname,name in dirCellHnames.items() if hname == sec.hname()})
secDic = {} synMechs = []

@@ -352,7 +352,7 @@ try: synMech[varName] = point.__getattribute__(varName) except:

print ‘Could not read variable %s from synapse %s’%(varName,synMech[‘label’])
print(‘Could not read variable %s from synapse %s’%(varName,synMech[‘label’]))
if not [_equal_dicts(synMech, synMech2, ignore_keys=[‘label’]) for synMech2 in synMechs]: synMechs.append(synMech)

@@ -368,7 +368,7 @@

except:

print ‘Could not read %s variable from point process %s’%(varName,pointpName)
print(‘Could not read %s variable from point process %s’%(varName,pointpName))
if pointps: secDic[secName][‘pointps’] = pointps

@@ -385,7 +385,7 @@ secLists = h.List(‘SectionList’) if int(secLists.count()): secListDic = {}

for i in xrange(int(secLists.count())): # loop over section lists
for i in range(int(secLists.count())): # loop over section lists hname = secLists.o(i).hname() if hname in dirCellHnames: # use python variable name secListName = dirCellHnames[hname]

@@ -396,9 +396,9 @@ secListDic = {}

globs = getGlobals(varList[‘mechs’].keys()+varList[‘pointps’].keys(), origGlob=origGlob)
globs = getGlobals(list(varList[‘mechs’].keys())+list(varList[‘pointps’].keys()), origGlob=origGlob) if ‘v_init’ in globs: # set v_init for each section (allows for cells with differnet vinits)
for sec in secDic.values(): sec[‘vinit’] = globs[‘v_init’]
for sec in list(secDic.values()): sec[‘vinit’] = globs[‘v_init’]
h.initnrn()

@@ -435,7 +435,7 @@ f.write(connText) # write starting text for row in range(1,numRows+1): if sheet.cell(row=row, column=colProb).value: # if not empty row

print ‘Creating conn rule for row ’ + str(row)
print(‘Creating conn rule for row ’ + str(row))
pre = sheet.cell(row=row, column=colPreTags).value post = sheet.cell(row=row, column=colPostTags).value

RefactoringTool: Writing converted netpyne/utils.py to netpyne_py3/utils.py. RefactoringTool: Refactored netpyne/wrappers.py — netpyne/wrappers.py (original) +++ netpyne/wrappers.py (refactored) @@ -11,7 +11,7 @@ __all__.extend([‘create’, ‘simulate’, ‘analyze’, ‘createSimulate’, ‘createSimulateAnalyze’, ‘load’, ‘loadSimulate’, ‘loadSimulateAnalyze’, \ ‘createExportNeuroML2’,’importNeuroML2SimulateAnalyze’]) # wrappers

-import sim +from . import sim

###############################################################################

RefactoringTool: Writing converted netpyne/wrappers.py to netpyne_py3/wrappers.py. RefactoringTool: No changes to netpyne/support/__init__.py RefactoringTool: Writing converted netpyne/support/__init__.py to netpyne_py3/support/__init__.py. RefactoringTool: Refactored netpyne/support/bsmart.py — netpyne/support/bsmart.py (original) +++ netpyne/support/bsmart.py (refactored) @@ -95,7 +95,7 @@ try: # First, try the Cholesky decomposition output=linalg.cholesky(M) except: # If not, just return garbage

print ‘WARNING: Cholesky failed, so returning (invalid) identity matrix!’
print(‘WARNING: Cholesky failed, so returning (invalid) identity matrix!’) output=matrix(eye(size(M,0)))
return output

@@ -234,7 +234,7 @@ [L,N] = shape(x); #L is the number of channels, N is the total points in every channel

if freq==0: F=timefreq(x[0,:],fs) # Define the frequency points

else: F=array(range(0,freq+1)) # Or just pick them
else: F=array(list(range(0,freq+1))) # Or just pick them npts=size(F,0)
maxindex=np.sum(arange(1,L))

@@ -298,11 +298,11 @@

Version: 2011jul18 “””

from bsmart import timefreq, pwcausalr
from .bsmart import timefreq, pwcausalr from scipy import array, size
if maxfreq==0: F=timefreq(vec1,rate) # Define the frequency points
else: F=array(range(0,maxfreq+1)) # Or just pick them
else: F=array(list(range(0,maxfreq+1))) # Or just pick them npts=size(F,0)
data=array([vec1,vec2])

RefactoringTool: Writing converted netpyne/support/bsmart.py to netpyne_py3/support/bsmart.py. RefactoringTool: Refactored netpyne/support/morphology.py — netpyne/support/morphology.py (original) +++ netpyne/support/morphology.py (refactored) @@ -1,6 +1,6 @@

-from __future__ import division + import numpy as np import pylab as plt from matplotlib.pyplot import cm @@ -80,7 +80,7 @@

swc_secs = i3d.swc.sections

swc_secs = [swc_secs.object(i) for i in xrange(int(swc_secs.count()))]
swc_secs = [swc_secs.object(i) for i in range(int(swc_secs.count()))]
sec_list = {1: cell.soma, 2: cell.axon, 3: cell.dend, 4: cell.apic}

@@ -113,7 +113,7 @@ swc_sec.raw.getval(2, 0), sec=sec)

j = swc_sec.first

xx, yy, zz = [swc_sec.raw.getrow(i).c(j) for i in xrange(3)]
xx, yy, zz = [swc_sec.raw.getrow(i).c(j) for i in range(3)] dd = swc_sec.d.c(j) if swc_sec.iscontour_:

@@ -554,10 +554,10 @@ ‘section_orientation’: h.section_orientation(sec=sec), ‘parent’: my_parent, ‘parent_loc’: my_parent_loc,

‘x’: [h.x3d(i, sec=sec) for i in xrange(n3d)],
‘y’: [h.y3d(i, sec=sec) for i in xrange(n3d)],
‘z’: [h.z3d(i, sec=sec) for i in xrange(n3d)],
‘diam’: [h.diam3d(i, sec=sec) for i in xrange(n3d)],
‘x’: [h.x3d(i, sec=sec) for i in range(n3d)],
‘y’: [h.y3d(i, sec=sec) for i in range(n3d)],
‘z’: [h.z3d(i, sec=sec) for i in range(n3d)],
‘diam’: [h.diam3d(i, sec=sec) for i in range(n3d)], ‘name’: sec.hname() })

RefactoringTool: Writing converted netpyne/support/morphology.py to netpyne_py3/support/morphology.py. RefactoringTool: Refactored netpyne/support/stackedBarGraph.py — netpyne/support/stackedBarGraph.py (original) +++ netpyne/support/stackedBarGraph.py (refactored) @@ -182,7 +182,7 @@ data_copy /= data_stack[levels-1] data_stack /= data_stack[levels-1] if heights is not None:

print “WARNING: setting scale and heights does not make sense.”
print(“WARNING: setting scale and heights does not make sense.”) heights = None elif heights is not None: data_copy /= data_stack[levels-1]

RefactoringTool: Writing converted netpyne/support/stackedBarGraph.py to netpyne_py3/support/stackedBarGraph.py. RefactoringTool: Files that were modified: RefactoringTool: netpyne/__init__.py RefactoringTool: netpyne/analysis.py RefactoringTool: netpyne/batch.py RefactoringTool: netpyne/cell.py RefactoringTool: netpyne/network.py RefactoringTool: netpyne/neuromlFuncs.py RefactoringTool: netpyne/pop.py RefactoringTool: netpyne/sim.py RefactoringTool: netpyne/simFuncs.py RefactoringTool: netpyne/specs.py RefactoringTool: netpyne/utils.py RefactoringTool: netpyne/wrappers.py RefactoringTool: netpyne/support/__init__.py RefactoringTool: netpyne/support/bsmart.py RefactoringTool: netpyne/support/morphology.py RefactoringTool: netpyne/support/stackedBarGraph.py Salvador-Duras-MacBook-Pro%

installing py3 to test

bash conda create -n py35 python=3.5 anaconda source activate py35 export PYTHONPATH=/usr/site/nrniv/local/python:/usr/arch/nrn/share/python/lib/python unset PYTHONHOME

(py35) bash-3.2$ python Python 3.5.2 |Anaconda 4.3.1 (x86_64)| (default, Jul 2 2016, 17:52:12) [GCC 4.2.1 Compatible Apple LLVM 4.2 (clang-425.0.28)] on darwin Type “help”, “copyright”, “credits” or “license” for more information. >>> from neuron import h Traceback (most recent call last): File “<stdin>”, line 1, in <module> File “/usr/arch/nrn/share/python/lib/python/neuron/__init__.py”, line 120 print “Enabling NEURON+Python help system.” ^ SyntaxError: Missing parentheses in call to ‘print’

sam: need to recompile Neuron and specify python3

17Jun05 Debugging netpyne (forum q)

Steps

netParams.py (error checking)
init.py

– error msgs (understand conceptual implementation) – verbose – inspect instanced net – results

Possible causes

netParams syntax,format,typo - error check should help
netParams conceptual (eg. label doesn’t exist) - error check
netpyne bug

17Oct17 Fixing issue importing cell

problem

Importing cell works in Penny’s computer but not mine or Sergios works = produces same outptu in NEURON/Py than in NetPyNE

same code (penny+me)

NEURON/Py code

from CA229_PFC import * import matplotlib.pyplot as plt from neuron import h import numpy as np #import json #import time #import pdb # For python debugging

cell = CA229() cell.init_once()

Stim = h.IClamp(cell.soma[2](0.5)) Stim.dur = 300 Stim.delay = 200 Stim.amp = 0.2

t_vec = h.Vector() t_vec.record(h._ref_t) v_vec_soma = h.Vector() v_vec_dend1 = h.Vector() v_vec_dend2 = h.Vector()

v_vec_soma.record(cell.soma[2](0.5)._ref_v) #v_vec_dend1.record(cell.basal[8](0.9)._ref_v) v_vec_dend2.record(cell.basal[34](0.8)._ref_v)

h.tstop = 500 h.v_init = -67.3 h.celsius = 32 h.run() # plot voltage vs time

plt.figure() # Default figsize is (8,6) plt.plot(t_vec, v_vec_soma, label = ‘Soma[0](0.5)’, color = ‘blue’) #plt.plot(t_vec, v_vec_dend1, label = ‘Basal[8](0.9)’, color = ‘green’) plt.plot(t_vec, v_vec_dend2, label = ‘Basal[34](0.8)’, color = ‘red’) plt.ylim([-80, 60]) plt.xlim([0,500]) plt.legend(loc = ‘best’) plt.show()

netpyne code

””” params.py

netParams is a dict containing a set of network parameters using a standardized structure

simConfig is a dict containing a set of simulation configurations using a standardized structure

Contributors: salvadordura@gmail.com “””

#import matplotlib; matplotlib.use(‘Agg’)

from netpyne import specs

netParams = specs.NetParams() # object of class NetParams to store the network parameters simConfig = specs.SimConfig() # object of class SimConfig to store the simulation configuration

#netParams.scaleConnWeightNetStims = 1.0 #0.5 # scale conn weight factor for NetStims ############################################################################### #

# ###############################################################################

###############################################################################

##########

cellRule = netParams.importCellParams(label=’PFC_full’,conds={‘cellType’: ‘PFC’, ‘cellModel’: ‘PFC_full’}, fileName=’CA229_PFC.py’, cellName=’MakeCA229’)

netParams.cellParams[‘PFC_cell_rule’] = cellRule

data = netParams.cellParams[‘PFC_cell_rule’]

#import json #with open(‘CellRule.json’,’w’) as output:

########## secName is the compartment name: soma_0, soma_1, basal_1, basal_2 #for secName,sec in cellRule[‘secs’].iteritems(): sec[‘v_init’] = -67.59

netParams.popParams[‘PFC’] = {‘cellType’: ‘PFC’, ‘cellModel’: ‘PFC_full’, ‘numCells’: 1}

#netParams.synMechParams[‘AMPA’] = {‘mod’:’GLU’} #netParams.synMechParams[‘NMDA’] = {‘mod’:’NMDA’,’Beta’: 0.01}

netParams.stimSourceParams[‘Stim’] = {‘type’:’IClamp’,’delay’:200, ‘amp’: 0.2, ‘dur’:300} netParams.stimTargetParams[‘Stim->PFC’] = {‘source’: ‘Stim’,’conds’: {‘pop’:’PFC’}, ‘sec’: ‘soma_2’, ‘loc’: 0.5} #netParams.popParams[‘ST1’] = {‘cellModel’: ‘NetStim’, ‘numCells’: 1, ‘rate’:50, ‘noise’: 0, ‘start’:50, ‘number’: 1} #netParams.popParams[‘ST1_1’] = {‘cellModel’: ‘NetStim’, ‘numCells’: 1, ‘rate’: 50, ‘noise’: 0, ‘start’:50, ‘number’: 1} #netParams.popParams[‘ST2’] = {‘cellModel’: ‘NetStim’, ‘numCells’: 1, ‘rate’:50, ‘noise’: 0, ‘start’:50, ‘number’: 1} #import itertools

#w_AMPA = list(itertools.repeat(0.00058, 20)) #loc_AMPA = list(itertools.repeat(0.6, 20)) ###### Connectivity parameters #netParams.connParams[‘ST1’] = {

#

#netParams.connParams[‘ST2’] = {

###############################################################################

#simConfig.hParams = {‘celsius’: 6} simConfig.duration = 500 # Duration of the simulation, in ms simConfig.dt = 0.025 # Internal integration timestep to use simConfig.seeds = {‘conn’: 1, ‘stim’: 1, ‘loc’: 1} # Seeds for randomizers (connectivity, input stimulation and cell locations) simConfig.createNEURONObj = 1 # create HOC objects when instantiating network simConfig.createPyStruct = 1 # create Python structure (simulator-independent) when instantiating network simConfig.verbose = False # show detailed messages simConfig.hParams = {‘celsius’: 32, ‘v_init’: -67.3}

#’celsius’: 32,

simConfig.recordCells = [‘all’] # which cells to record from simConfig.recordTraces = {‘V_soma’:{‘sec’:’soma_2’,’loc’:0.5,’var’:’v’}} simConfig.recordTraces[‘Bdend1’] = {‘sec’:’basal_34’, ‘loc’: 0.8, ‘var’: ‘v’} simConfig.recordTraces[‘Bdend2’] = {‘sec’:’basal_34’, ‘loc’: 0.5, ‘var’: ‘v’} simConfig.recordTraces[‘Bdend3’] = {‘sec’:’basal_34’, ‘loc’: 0.3, ‘var’: ‘v’} #simConfig.recordTraces[‘Bdend’] = {‘sec’:’basal_34’, ‘loc’: 0.8, ‘var’: ‘v’} simConfig.recordStim = False # record spikes of cell stims simConfig.recordStep = 0.1 # Step size in ms to save data (eg. V traces, LFP, etc)

simConfig.filename = ‘HHTut’ # Set file output name #simConfig.Label = ‘sim1’ simConfig.saveFileStep = 1000 # step size in ms to save data to disk simConfig.savePickle = False # Whether or not to write spikes etc. to a .mat file simConfig.recordStim = True simConfig.saveJson = True # save to json file

simConfig.analysis[‘plotRaster’] = False # Plot raster simConfig.analysis[‘plotTraces’] = {‘include’: [0], ‘overlay’: True, ‘saveFig’: True, ‘showFig’: False} # Plot raster simConfig.analysis[‘plot2Dnet’] = False # Plot 2D net cells and connections

from netpyne import sim sim.createSimulateAnalyze(netParams = netParams, simConfig = simConfig)

import pylab; pylab.show()

same/similar NEURON version

penny: NEURON – VERSION 7.5 master (8c17845) 2017-06-22 me: NEURON – VERSION 7.5 master (0388d94) 2017-08-09

python version

penny: Python 2.7.13 |Continuum Analytics, Inc.| (default, Dec 20 2016, 23:05:08) me: Python 2.7.13 |Anaconda custom (x86_64)| (default, Dec 20 2016, 23:05:08)

netpyne version!

difference in netpyne: sim.replaceDictODict()

due to an obscure bug resulting from using `isinstance()` instead of `type()` (isinstance takes into account inherited

classes, so was applied to wrong objects and messed up imported structures)

17Oct18 Implementing HNN’s dipole

chat with sam

samn [8:58 AM] looks like two components dipole.mod (non point process with pointer) dipole_pp.mod (point process)

salvadord [11:48 AM] not familiar with non-point process with pointer — any doc/notes on that ?

samn [11:49 AM] not much - stephanie developed that with michael hines a long time ago and then shane lee translated it to python with michael’s help

[11:49] maybe i should email steph/michael and cc you

[11:49] dipole_insert in cell.py has a few comments

salvadord [11:50 AM] I’ll check the code and if can’t figure out will let u know

samn [11:50 AM] mainly what dipole needs is the axial resistance and voltage in different locations so that it can calculate the axial current flow

salvadord [11:50 AM] ic

samn [11:51 AM] wonder if mh could add something for getting axial current in an easier way; similar to new feature of transmembrane current

17Oct18 Automating conversion from py2 to py3

steps required and which can be included in py2

2to3

convert everything, including examples and docs

- replace global import sim with local in each func

can probably be done now in py2

fix import neuromlFuncs and analysis

use py3 format

from . import x import .x

id32() encode utf-8

use in py2 (error?)

return int(hashlib.md5(obj.encode(‘utf-8’)).hexdigest()[0:8],16)

replace popen2 with Popen in batch.py

not sure if Popen available in py2?

py2: output, input = popen2(‘qsub’) # Open a pipe to the qsub command.

py3: proc = Popen(command.split(’ ‘), stdin=PIPE, stdout=PIPE) # Open a pipe to the qsub command. (output, input) = (proc.stdin, proc.stdout)

replaced exec to import with importlib.import_module

test in py2 - looks like more elegant solution
import importlib tempModule = importlib.import_module(moduleName)

replaced x.setattr(…) with setattr(x,…)

4 occur in cell.py - more stable, use in py2

create script to convert py2 to 3

from subprocess import call

py2_root = ‘../netpyne_repo’ folders = [‘netpyne’, ‘doc’, ‘examples’] files = [‘CHANGES.md’, ‘README.md’, ‘sdnotes.org’, ‘.gitignore’]

for file in files: call((‘rm %s’%(file)).split(’ ‘)) call((‘cp %s/%s .’%(py2_root, file)).split(’ ‘))

for folder in folders: call((‘rm -r %s’%(folder)).split(’ ‘)) call((‘2to3 –output-dir=./%s -W -n %s/%s’%(folder, py2_root, folder)).split(’ ‘))

error checking error in py3

Traceback (most recent call last): File “/usr/site/python3/netpyne/netpyne/tests/tests.py”, line 2849, in execRunTests self.testTypeObj.testIsBoolean(paramName) File “/usr/site/python3/netpyne/netpyne/tests/tests.py”, line 257, in testIsBoolean assert (isinstance (val,bool) or val in [0,1]), “Value specified is ” + str(val) + “.” AssertionError: Value specified is 0.1.

During handling of the above exception, another exception occurred:

Traceback (most recent call last): File “init.py”, line 8, in <module> sim.createSimulateAnalyze(simConfig = cfg, netParams = netParams) File “/usr/site/python3/netpyne/netpyne/wrappers.py”, line 73, in createSimulateAnalyze (pops, cells, conns, stims, simData) = sim.create(netParams, simConfig, output=True) File “/usr/site/python3/netpyne/netpyne/wrappers.py”, line 25, in create sim.initialize(netParams, simConfig) # create network object and set cfg and net params File “/usr/site/python3/netpyne/netpyne/simFuncs.py”, line 72, in initialize simTestObj.runTests() File “/usr/site/python3/netpyne/netpyne/tests/tests.py”, line 1520, in runTests self.runSimConfigTests() # load simConfig tests File “/usr/site/python3/netpyne/netpyne/tests/tests.py”, line 2465, in runSimConfigTests self.execRunTests(simConfigTestObj, simConfigParams) File “/usr/site/python3/netpyne/netpyne/tests/tests.py”, line 2857, in execRunTests print(str(testObj.errorMessageLevel[testIndex]) + “: ” + str(testObj.messageText[testIndex]) + str(e[0])) TypeError: ‘AssertionError’ object is not subscriptable

17Nov01 Issue reproducing PT cell results

problem

when run vs load from saved get different result
also in old vs new version of netpyne where the way to convert netParams to dict changed
saved netParams and net in json are identical in good vs bad version

possible explanations/solutions

order of mechs saved (order of instnatiation affects since dependent global variables?)
global variables not saved properly

solved

was due to using getattr() instead of x.get() when setting global variables

17Nov22 MPI error during gathering

forum link

https://www.neuron.yale.edu/phpBB/viewtopic.php?f=45&t=3770&sid=7945c6bfcb495b22a4b3e72c0850d885

testing

tried on zn

24 nodes gave error
trying with:

simConfig.gatherOnlySimData = True simConfig.saveCellSecs = False simConfig.saveCellConns = False

only a couple traces actually recorded so small simData – maybe bug because empty? WORKED OK!

17Dec21 GUI

setup

checkout and pull all branches:

Pavos-MBP% python gitall.py branches NEURON_UI development

org.geppetto.frontend.jupyter development

org.geppetto.frontend interface

Geppetto Neuron interface

npm install from /u/salvadord/Documents/ISB/Models/NEURON-UI/org.geppetto.frontend.jupyter/src/jupyter_geppetto/geppetto/src/main/webapp
in utilities python run-dev-server.py
checkout and pull metadata branch in netpyne repo

Extend GUI

In browser open console + disable cache (eg. Network -> “Disable cache” with browser open) – so changes in js take effect after reloading if run-dev-server.py running
using javascript REACT – component based
start in:

/u/salvadord/Documents/ISB/Models/NEURON-UI/org.geppetto.frontend.jupyter/src/jupyter_geppetto/geppetto/src/main/webapp/extensions/geppetto-neuron

render() method
check imports for other components

to check can import tut example

from neuron_ui.tests.tut3 import netParams, simConfig netpyne_geppetto.netParams=netParams netpyne_geppetto.simConfig=simConfig neuron-ui

create branch and then PR

set background color

CanvasContainer.setBackgroundColor(‘#000’)

set cell cylinders

network.PT5B[0].setGeometryType(‘cylinders’)

17Dec22 Extracellular recording (LFP) from Allen BMTK Bionet

recxelectrode class

https://github.com/AllenInstitute/bmtk/blob/develop/bmtk/simulator/bionet/recxelectrode.py

simulation code

https://github.com/AllenInstitute/bmtk/blob/a20b10af77563e2dbebb8c2db7c3c8878783d0f9/bmtk/simulator/bionet/simulation.py

pre sim (deprecated?)

def calculate_ecp(self, positions_file, ecp_output): “”“Set recording of the ExtraCellular Potential :param positions_file: “”” self._ecp_output = ecp_output self._calc_ecp = True self.rel = RecXElectrode(positions_file) for gid in self.gids[‘biophysical’]: cell = self.net.cells[gid] self.rel.calc_transfer_resistance(gid, cell.get_seg_coords())

self._rel_nsites = self.rel.nsites

h.cvode.use_fast_imem(1) # make i_membrane_ a range variable self.fih1 = h.FInitializeHandler(0, self.set_pointers)

compute and save after sim

if self._calc_ecp: nsites = self.rel.nsites self.data_block[‘ecp’] = np.empty((nt_block, nsites))

if self._calc_ecp: for gid in self.gids[‘biophysical’]: # compute ecp only from the biophysical cells cell = self.net.cells[gid] im = cell.get_im() tr = self.rel.get_transfer_resistance(gid) ecp = np.dot(tr, im)

if self.cell_variables and gid in self.gids[‘save_cell_vars’]: cell_data_block = self.data_block[‘cells’][gid] cell_data_block[‘ecp’][tstep_block-1, :] = ecp

self.data_block[‘ecp’][tstep_block-1, :] += ecp

if calc_ecp: self.im_ptr = h.PtrVector(self._nseg) # pointer vector self.im_ptr.ptr_update_callback(self.set_im_ptr) # used for gathering an array of i_membrane values from the pointer vector self.imVec = h.Vector(self._nseg)

def set_im_ptr(self): “”“Set PtrVector to point to the i_membrane_”“” jseg = 0 for sec in self.hobj.all: for seg in sec: self.im_ptr.pset(jseg,seg._ref_i_membrane_) # notice the underscore at the end jseg += 1

cell methods required

def calc_seg_coords(self, morph_seg_coords): “”“Calculate segment coordinates for individual cells”“” phi_y = self._props.rotation_angle_yaxis # self._props[‘rotation_angle_yaxis’] phi_z = self._props.rotation_angle_zaxis # self._props[‘rotation_angle_zaxis’]

RotY = utils.rotation_matrix([0, 1, 0], phi_y) # rotate segments around yaxis normal to pia RotZ = utils.rotation_matrix([0, 0, 1], -phi_z) # rotate segments around zaxis to get a proper orientation RotYZ = RotY.dot(RotZ)

self._seg_coords[‘p0’] = self._pos_soma + np.dot(RotYZ, morph_seg_coords[‘p0’]) self._seg_coords[‘p1’] = self._pos_soma + np.dot(RotYZ, morph_seg_coords[‘p1’])

def get_seg_coords(self): return self._seg_coords

def get_im(self): “”“Gather membrane currents from PtrVector into imVec (does not need a loop!)”“” self.im_ptr.gather(self.imVec) return self.imVec.as_numpy() # (nA)

insights

seg_coords[‘p0’] = array with the starting 3D point of each cell segment seg_coords[‘p1’] = array with the ending 3D point of each cell segment

implementation

specs.py

cfg.recordECP = True

cell.py

if cfg.recordECP: self.im_ptr = h.PtrVector(self._nseg) # pointer vector self.im_ptr.ptr_update_callback(self.set_im_ptr) # used for gathering an array of i_membrane values from the pointer vector

def set_im_ptr(self): “”“Set PtrVector to point to the i_membrane_”“” jseg = 0 for sec in self.hobj.all: for seg in sec: self.im_ptr.pset(jseg,seg._ref_i_membrane_) # notice the underscore at the end jseg += 1

network.py (or simFuncs.py?)

self.calc_seg_coords() # use for computing the ECP

def calc_seg_coords(self): – call method in each cell “”“Needed for the ECP calculations”“” for node_type_id, morphology in self.__morphologies_cache.items(): morph_seg_coords = morphology.calc_seg_coords() # needed for ECP calculations

for node in self._local_node_types[node_type_id]: self._cells[node.node_id].calc_seg_coords(morph_seg_coords)

calc seg coords for each cell class (relative to each morphology, so don’t repeat) - calculate for 1 cell of each pop
adapt to each cell by translating by x,y,z

simFuncs

if cfg.recordECP: set_ecp_recording()

def set_ecp_recording(self):

”’ Set recording of the ExtraCellular Potential ”’

self.rel = RecXElectrode(self.conf)

for gid in self.gids[‘biophysical’]: cell = self.net.cells[gid] self.rel.calc_transfer_resistance(gid, cell.get_seg_coords())

h.cvode.use_fast_imem(1) # make i_membrane_ a range variable self.fih1 = h.FInitializeHandler(0, self.set_pointers)

def set_pointers(self): # set pointers to i_membrane in each cell for gid, cell in self.net.cells.items(): cell.set_im_ptr()

calcualte ECP: for gid in self.gids[‘biophysical’]: # compute ecp only from the biophysical cells
cell = self.net.cells[gid] im = cell.get_im() tr = self.rel.get_transfer_resistance(gid) ecp = np.dot(tr,im)

if self.conf[‘run’][‘save_cell_vars’] and gid in self.gids[‘save_cell_vars’] : cell_data_block = self.data_block[‘cells’][gid] cell_data_block[‘ecp’][tstep_block-1,:] = ecp

self.data_block[‘ecp’][tstep_block-1,:] += ecp

store ECP in simData

17Dec23 LFP using Sam’s simple method

classes

from math import sqrt,pi

class currsource: def __init__ (self, ptr, x, y, z): self.ptr = ptr self.x = x self.y = y self.z = z def getcurr (self, elx, ely, elz): d = sqrt((elx - self.x)**2 + (ely - self.y)**2 + (elz - self.z)**2) if d > 0: val = self.ptr[0] return val / d return 0.0 def getcurrd (self, d): if d > 0.0: return self.ptr[0] / d return 0.0

class pointlfp: def __init__ (self, elx, ely, elz, lcurr = [], sigma=1.0): self.lcurr = lcurr self.elx = elx self.ely = ely self.elz = elz self.sigma = 1.0 self.vlfp = [] self.ld = [] def setld (self): for currs in self.lcurr: self.ld.append(sqrt((self.elx - currs.x)**2 + (self.ely - currs.y)**2 + (self.elz - currs.z)**2)) def calcv (self): if len(self.ld) < 1: self.setld() i = 0.0 for currs,d in zip(self.lcurr,self.ld): i += currs.getcurrd(d) return i / (4.0*pi*self.sigma) def reset (self): self.vlfp = [] self.ld = [] def savev (self): self.vlfp.append(self.calcv())

main network code

def pyrcurrsrc (cell,lcurrsrc): for sec in [cell.Bdend, cell.soma, cell.Adend1, cell.Adend2, cell.Adend3]: lcurrsrc.append(currsource(sec(0.5).pas._ref_i, cell.x, cell.y, getzpos(cell,sec))) lcurrsrc.append(currsource(sec(0.5).nacurrent._ref_ina, cell.x, cell.y, getzpos(cell,sec))) lcurrsrc.append(currsource(sec(0.5).kacurrent._ref_ika, cell.x, cell.y, getzpos(cell,sec))) lcurrsrc.append(currsource(sec(0.5).kdrcurrent._ref_ik, cell.x, cell.y, getzpos(cell,sec))) lcurrsrc.append(currsource(sec(0.5).hcurrent._ref_ih, cell.x, cell.y, getzpos(cell,sec))) lcurrsrc.append(currsource(sec(0.5)._ref_i_cap, cell.x, cell.y, getzpos(cell,sec))) lcurrsrc.append(currsource(cell.somaGABAf.syn._ref_i, cell.x, cell.y, getzpos(cell,cell.soma))) lcurrsrc.append(currsource(cell.somaAMPAf.syn._ref_i, cell.x, cell.y, getzpos(cell,cell.soma))) lcurrsrc.append(currsource(cell.BdendAMPA.syn._ref_i, cell.x, cell.y, getzpos(cell,cell.Bdend))) lcurrsrc.append(currsource(cell.BdendNMDA.syn._ref_iNMDA, cell.x, cell.y, getzpos(cell,cell.Bdend))) lcurrsrc.append(currsource(cell.BdendNMDA.syn._ref_ica, cell.x, cell.y, getzpos(cell,cell.Bdend))) lcurrsrc.append(currsource(cell.Adend3GABAs.syn._ref_i, cell.x, cell.y, getzpos(cell,cell.Adend3))) lcurrsrc.append(currsource(cell.Adend3GABAf.syn._ref_i, cell.x, cell.y, getzpos(cell,cell.Adend3))) lcurrsrc.append(currsource(cell.Adend3NMDA.syn._ref_iNMDA, cell.x, cell.y, getzpos(cell,cell.Adend3))) lcurrsrc.append(currsource(cell.Adend3NMDA.syn._ref_ica, cell.x, cell.y, getzpos(cell,cell.Adend3))) lcurrsrc.append(currsource(cell.Adend3AMPAf.syn._ref_i, cell.x, cell.y, getzpos(cell,cell.Adend3)))

def allpyrcurrsrc (): lcurrsrc = [] for cell in net.pyr.cell: pyrcurrsrc(cell,lcurrsrc) return lcurrsrc

def makepflp (elx,ely,elz): plfp = pointlfp(elx,ely,elz,allpyrcurrsrc()) return plfp

def recpyrcurr (cell,lrec,lx,ly,lz): xpos,ypos = cell.x,cell.y; for sec in [cell.Bdend, cell.soma, cell.Adend1, cell.Adend2, cell.Adend3]: reccurr(lrec,sec(0.5).pas._ref_i); savepos(lx,ly,lz,cell,sec) reccurr(lrec,sec(0.5).nacurrent._ref_ina); savepos(lx,ly,lz,cell,sec) reccurr(lrec,sec(0.5).kacurrent._ref_ika); savepos(lx,ly,lz,cell,sec) reccurr(lrec,sec(0.5).kdrcurrent._ref_ik); savepos(lx,ly,lz,cell,sec) reccurr(lrec,sec(0.5).hcurrent._ref_ih); savepos(lx,ly,lz,cell,sec) reccurr(lrec,sec(0.5)._ref_i_cap); savepos(lx,ly,lz,cell,sec) reccurr(lrec,cell.somaGABAf.syn._ref_i); savepos(lx,ly,lz,cell,cell.soma) reccurr(lrec,cell.somaAMPAf.syn._ref_i); savepos(lx,ly,lz,cell,cell.soma) reccurr(lrec,cell.BdendAMPA.syn._ref_i); savepos(lx,ly,lz,cell,cell.Bdend) reccurr(lrec,cell.BdendNMDA.syn._ref_iNMDA); savepos(lx,ly,lz,cell,cell.Bdend) reccurr(lrec,cell.BdendNMDA.syn._ref_ica); savepos(lx,ly,lz,cell,cell.Bdend) reccurr(lrec,cell.Adend3GABAs.syn._ref_i); savepos(lx,ly,lz,cell,cell.Adend3) reccurr(lrec,cell.Adend3GABAf.syn._ref_i); savepos(lx,ly,lz,cell,cell.Adend3) reccurr(lrec,cell.Adend3NMDA.syn._ref_iNMDA); savepos(lx,ly,lz,cell,cell.Adend3) reccurr(lrec,cell.Adend3NMDA.syn._ref_ica); savepos(lx,ly,lz,cell,cell.Adend3) reccurr(lrec,cell.Adend3AMPAf.syn._ref_i); savepos(lx,ly,lz,cell,cell.Adend3) return lrec,lx,ly,lz

def recallpyrcurr (): lrec,lx,ly,lz = [],[],[],[] for cell in net.pyr.cell: recpyrcurr(cell,lrec,lx,ly,lz) return lrec,lx,ly,lz

def genLFP (lrec,lx,ly,lz,elx,ely,elz): vlfp = h.Vector(lrec[0].size()); vtmp=h.Vector(lrec[0].size()) for i,vec in enumerate(lrec): vtmp.copy(vec) d = sqrt((elx - lx[i])**2 + (ely - ly[i])**2 + (elz - lz[i])**2) if d > 0: vtmp.mul(1.0/d) vlfp.add(vtmp) sigma = 1.0 # arbitrary vlfp.mul(1.0/(4.0*pi*sigma)) return vlfp

17Dec02 Implementing RxD

celParams data model (inside cellParams)

cellParams[‘rule1’] (dict)

– [‘rxd’] (dict) — [‘Species’] (dict) ---- [label] (dict) ----- [‘d’: , ‘name’: , ‘charge’: ,…] — [‘Region’] (dict) ---- [label] (dict) ----- [‘secs’: ] — [‘Reaction’] (dict) — [‘Rate’] (dict) — [‘MultiCompartmentReaction’] (dict) — [‘Extracellular’] (dict)

cell instance data model (inside cell)

Cell (object)

– rxd (dict) — [‘Species’] (dict) ---- [label] (dict) ----- [‘d’: , ‘name’: , ‘charge’: ,…] — [‘Region’] (dict) ---- [label] (dict) ----- [‘secs’: ] — [‘Reaction’] (dict) — [‘Rate’] (dict) — [‘MultiCompartmentReaction’] (dict) — [‘Extracellular’] (dict)

rxdParams data model (new attribute of netParams)

rxdParams (dict)

– [‘Species’] (dict) — [label] (dict) ---- [‘d’: , ‘name’: , ‘charge’: ,…] – [‘Region’] (dict) — [label] (dict) ---- [‘secs’: ] % conditions to define cells that sections apply to – [‘Reaction’] (dict) – [‘Rate’] (dict) – [‘MultiCompartmentReaction’] (dict) – [‘Extracellular’] (dict)

Steps to release new version

Update CHANGES.md
/netpyne/__init__.py __version__=’0.7.8’
gcp “VERSION 0.7.8” (from development branch)
create PR from development to master (via https://github.com/Neurosim-lab/netpyne)
make sure no conflicts and Travis tests pass
merge PR
“bnp” alias (to build documentation)
“npr” alias(to release on pip)
Announce new release:

– on github: https://github.com/Neurosim-lab/netpyne/releases – on NEURON forum: https://www.neuron.yale.edu/phpBB/viewtopic.php?f=45&t=3685&sid=9c380fe3a835babd47148c81ae71343e – on Google group: https://groups.google.com/forum/#!forum/netpyne-mailing – on Slack #netpyne – on Twitter (salvador account; create netpyne account?)

Netpyne Models/Users

Github examples folder

Documentation tutorials

M1 network (Salva, Neurosim)

Parallel Neuron paper (Alex, Neurosim)

Claustrum network (Jing, George Augustine)

Sensorimotor network + arm (Cliff, Kerr lab)

Hippocampus model (Ernie, Neurosim)

OSB/NeuroML (Padraig UCL)

V1 load/save model (Allen Brain Institute)

STDP (Anatoly Buchin, Allen)

STDP in networks of biophysically detailed cells

Basal Ganglia (Lucas, UCD)

http://www.neuromuscularsystemsucd.info/research-areas

email

I made a new pull request into the development branch as requested. I am using Netpyne to model the Basal Ganglia network using biophysically detailed and simplified spiking models. I am going to look at the effect of pathological oscillations on task performance for motor tasks. One of the overarching goals of our group is to combine different models into a multi-scale model of Basal Ganglia, spinal chord and motor units in the muscles, so I may use Netpyne for that too. I appreciate all the work that is made open source by your group. It helped me a lot while learning how to use the more advanced features of Neuron using Python. All the best, Lucas

LFP oscillations (Fink, Ohio Wesleyan)

Christian G. Fink, PhD Assistant Professor of Physics and Neuroscience Ohio Wesleyan University

Learning, dendritic computations (Birgit Kriener, Einevoll lab, University of Oslo)

Thalamocortical epilespy (Andrew Knox, Cincinatti)

Cortex, Schizophrenia (Cristoph Metzner, Hertfordshire)

Spinal cord circuits (Vittorio Caggiano, IBM Watson)

Convert hoc models (Ben Latimer, Missouri)

Thalamocortical stimulation (Brenyn Jungmann, Nahir lab,Missouri)

Fear circuits (Aliyah Taylor, Princeton/Missouri)

Macaque M1 TMS/tDCS (Aman Aberra, Duke)

CA1/CA3 (Mariane Bezaire, Boston University; Andras Ecker, HBP)

EEG (Sam Neymotin, Boston)

Cerebellum (Sergio Solinas)

Cerebellum (Stefano Masoli, Univ Pavia)

LFP in SUA vs MUA (Harry Tran) harrytran75005@gmail.com

the relationship between the single unit activty (SUA) and the multi unit activity (MUA) - so the LFP produced by a single neuron and its LFP in a population of neurons

EEE network and single cell (JOe, Sergio)

Ischemia with RxD

EMAIL LIST

salvadordura@gmail.com billl@neurosim.downstate.edu samnemo@gmail.com cliff@thekerrlab.com btessler92@gmail.com joe.w.graham@gmail.com james.chen@downstate.edu ahseidenstein@gmail.com andrew.knox@cchmc.org penggao.1987@gmail.com ramazan.tekin@batman.edu.tr sergio.angulo@gmail.com mitra.siddhartha@gmail.com adrianquintana@gmail.com matteo.cantarelli@gmail.com foxy1987@gmail.com christoph.metzner@googlemail.com mdistasio@gmail.com caggiano@gmail.com jing.lim.12@ucl.ac.uk davidkedz@hotmail.com p.gleeson@ucl.ac.uk anat.buchin@gmail.com lucas.koelman@ucdconnect.ie cgfink@owu.edu birgit.kriener@medisin.uio.no LatimerB@missouri.edu bwjmbb@mail.missouri.edu ali.taylor331@gmail.com aman.aberra@duke.edu marianne.bezaire@gmail.com smgsolinas@gmail.com bremen@fastwebnet.it harrytran75005@gmail.com

Related tools

NEURON

Multiscale simulator http://neuron.yale.edu/neuron/ \cite{Carn06,Lytt16}

Genesis

Multiscale simulator http://genesis-sim.org/ \cite{Bowe12}

MOOSE

Multiscale simulator https://moose.ncbs.res.in/ \cite{RayB08}

NEST

Large scale network simulator http://www.nest-simulator.org/ \cite{Ples15}

Nengo

Large scale network simulator http://www.nengo.ca/ \cite{Stew09}

GeNN

Large scale network simulator (GPUs) http://genn-team.github.io/genn/ \cite{Yavu16}

Brian

Mathematically oriented simulator http://briansimulator.org/ \cite{Good08}

MUSIC

Multi-simulator tool http://software.incf.org/software/music \cite{Djur10}

PyNN

Multi-simulator tool http://neuralensemble.org/PyNN/ \cite{Davi09}

NeuroConstruct

Multi-simulator tool http://www.neuroconstruct.org/ \cite{Glee07}

N2A

Multi-simulator tool https://github.com/frothga/n2a \cite{Roth14}

PyNeuroML/libNeuroML

Multi-simulator tool (for NeuroML2/LEMS models; combines libNeuroML, PyLEMS, jNeuroML) https://github.com/NeuroML/pyNeuroML \cite{Vell14}

Topographica

Multi-simulator tool (focused on topographic maps) http://ioam.github.io/topographica/ \cite{Bedn09}

NetPyNE

Tool for design/simulation/analysis of multiscale networks in NEURON https://neurosimlab.org/netpyne \cite{Dura16c}

Morphforge

Tool for design/simulation/analysis of single cell and simple nets in NEURON \cite{Hull14} https://morphforge.readthedocs.io

AllenSDK

Tool for design/simulation/analysis of single cell and simple nets in NEURON https://alleninstitute.github.io/AllenSDK/biophysical_models.html

Mozaik

Multi-simulator tool; Batch simulation organization and analysis http://neuralensemble.org/docs/mozaik/index.html \cite{Anto13}

Lancet

Batch simulation organization and analysis http://ioam.github.io/lancet/ \cite{Stev13}

SimTracker

Batch simulation organization and analysis https://bitbucket.org/mbezaire/simtrackercode http://mariannebezaire.com/simtracker/ \cite{} - under review

Sumatra

Batch simulation organization and analysis http://sumatra.readthedocs.io/ \cite{Dav14}

pypet

Parameter optimization; batch simulation organization and analysis https://pypet.readthedocs.io \cite{Meye16}

BluePyOpt / HBP Collab GUI

Parameter optimization https://github.com/BlueBrain/BluePyOpt \cite{VanG16}

NeuroTune

Parameter optimization http://optimal-neuron.readthedocs.io/en/latest/ Mike Vella and Padraig Gleeson

Optimizer

Parameter optimization http://optimizer.readthedocs.io/en/latest/ \cite{Frie14b}

NeuronVisio

Analysis and visualization http://docs.neuronvisio.org/ \cite{Matt12}

Geppetto/OSBViewer

Analysis and visualization http://www.geppetto.org/ \cite{Glee15}

NeuroML

Declarative format for neural models https://www.neuroml.org/ \cite{Glee10}

NineML

Declarative format for neural models http://incf.github.io/nineml/ \cite{Raik12}

Reaction-diffusion simulators

RxD
MCell
Smoldyn
STEPS
NeuroRD
PSICS

Reviews / Other lists

\cite{Bret07}
\cite{McDo16}

http://software.incf.org/software/?getTopics=Computational%20neuroscience&b_start:int=0&-C= http://software.incf.org/software?getTopics=Large%20scale%20modeling&-C=

Paper

Title

NetPyNE: a tool to develop, simulate and analyze data-driven biological neuronal network models

Journal

PLOS Comp Biol (similar to X-J Wang paper)

Figures

overview of netpyne stages
structure of netParams
how netParams specs translate to network: a) pops with yfrac, b) cell Model HH (import) vs Izhi, b) conn rules (eg. cell type, yfrac, pop)
yfrac-based connectivity illustration - with real data
sCRACM-based subcellular example - with real data
simulation process - distribution of cells and gathering (cite Neural Comp)
RL+STDP … for behavioral task: arm (intervalUpdateFunc)
Analysis: raster, hist, LFP, 2dnet, conn matrix
Conversion to NeuroML: example of neuroml code? Gepetto/OSB visualization?
Example nets: M1, claustrum, RL

Grant (moved to grant folder /u/salvadord/Work/netpyne_grant/sdnotes.org)

16aug17 11:00 MBL bill discuss netpyne with hines - sald notesi []

value of dictionaries would be a small lambda function that gives generality
two things i’d like to see – notation is reasonable but very descriptive

– seems there would be a better way to organize it [similar to NeuroML but shorter; needs to be standardized format] – what can be additional value added for the user [easy creation, parallel sim and analysis of net; complex conn patterns simplified; subcellular; inputs]

– ultimate would the the traub model – very compact network level description of the traub model 13 cell types, ~128 projection types – 128 statements – shouldn’t have to write down the same things – pij, weight, delay (cf NQS) [conn matrix can be converted to netpynec conn rules eg. M1; but need to write down same things so have consistent standard format for all nets, eg. some not based on pij, dist-dep, yfrac] – extract all this from the current 15k line code [neuroml version can extract net instance; but will have to check manually for high-level conn rules]

lambda function style encourages compactness; or def if used more than once [can use; string-based more compact and same effect; complexity of lambda: have to include arguments (eg. dist, x loc), not same for all rules]
on the setup set – instantiation – target cells for a particular rank – only setup on that rank due to the sources, building the netcon depends on location of the source; rules have SRC->TARG info so need to use the net-rendezvous-rank strategy (a common idiom used to resolve these problems)
net-rendezvous-rank strategy: used if trying to decide if something exists or not; can’t answer question unless all of the types exist but they exist randomly distributed across all the ranks so have to gather info from all the ranks; were do we meet? – we don’t meet at a single master; we are in double loops (for targ over TARG for src over SRC) – here’s the answer: we rendezvous at target_gid%nhost to decide [can implement in netpyne]
bad (but still common) idiom: sometimes useful to do net-rendezvous-rank in 2 phases - build round-robin system first and then throw away and rebuild with balanced distribution
I mentioned mark hereld example of an interactive HPC (also cf vslice) sim that is managed thru a remote GUI [netpyne GUI, OSB gepetto, new NEURON GUI, NSG]
change name? – NetPyNe sounds to much like PYNN – mh also works with Andrew Davison so would like to avoid the apparent

overlap; domain of PYNN – metasimulator to construct models in NEURON, NEST, BRIAN, using neuromorophic simulators (spinnaker, analog VLSI); NeuroML, MUSIC (multi-simulator-coordinator == lower level data interchange) in somewhat similar domains [mostly phonetic similarity; many python packages have py so similar; discuss with Andrew in Janelia]

look into Nengo – language used to put together Eliasmith – prob has a good network connectivity algorithm/layout https://pythonhosted.org/nengo/examples/network_design.html: ‘def motor_cortex(command_threshold, n_neurons_per_command=30)’ https://pythonhosted.org/nengo/examples/network_design_advanced.html?highlight=tips%20tricks [included most of NEST conn features + others; Nengo very artificial non-bio eg. implement math funcs using spiking neurons]
separate issues: advantages of branch failure; cf Paul Rhodes branch point failure [?]

16aug26 MBL meeting (sald notes)

Reverse engineering

inverse problem: approximation, computer-aided design
use GA to aid in the inverse problem
netpyne can generate initial attempt to inverse network; then user can update the randomizer functions etc based on the source code
modelview gets lowest hangin fruit, but netpyne could do more sophisticated analysis to extract features
network

– find cell types – connections to other types etc – given projection is there a constant/pattern that can be reverse engineered, convergence

Try out example of reverse engineering – maybe CA3 model
reverse engineer data - eg. M1 model have physio + conns and turn into model
netpyne can import cells from hoc template or cell class instance
people have done before – create model that generates data, and from data use Bayesian learning to infer params
1000 noisy channels, get data and do experiments to get back params of sodium channels back
eg. get maximal conductance param from data - maximum likelihood (won’t be same since noise)
hidden parameters that have to be informed - markov model
process noise + experimental noise - log likelihood
always wrong - only approximation
3 approaches:

– 1) current modeldb - get approx same rule – 2) GA - get different rule from original – 3) BRAIN (reality)

Evolutionary

What to optimize

– 1) GA - Learning params (eg. learning rate) – 2) High-level specs (eg. prob of conn) – 3) Network instance params (eg. explicit weights)

What to invert from GA generated network

– 1) High-level specs (eg. prob of conn) – 2) Dynamical properties (can be used to fit)

Aim 1

Brett et al

grant approach

we have done this in the last 5 years - can we do this for the next 5 years
significance
innovation
investigator

redefine aims

Aim 1: front end

– GUI – dendritic connectivity

Aim 2: back end

– rendezvous

not good - front vs back
make Aim 2: parallelization

– parallelize setup - rendezvous – load balance – parameter optimization (evol etc)

subworlds

set of nodes working on same networks
can have 10 subworlds, each for 1 net
bulletin board - old 1994 - currently master but could be done as distributed bulletin baoard

GUI

interface to create network
simtracker managing of sims
simtracker summary of model data

16aug26 Bill notes

get output of modelview to read the flat instantiated model but currently doesn’t do networks well

Aim3 modelview outputs a small subset of cells (sees different templated / ?classes) and can export to netpyne

finds commonality also by common parameters, morphologies random values in 1 location – can’t currently find it – could find distributions but can’t find random seed (a cryptographic solution) does the netpyne and original create the same net – Sub## are two networks identical Sub## netpyne description of the flat file – straightforward subroutines for inverting random distros — eg sampling without replacement – if pick 50 out of 1000

can apply GA to the high level expression –

NetPyNe expresses it and modelview interprets it

LEMS, LEMD (low entropy model description)

current traub 20k fortran model to 20k hoc model

often have to factor out what is different among very similar fragments

should we start with the single cell? – could that be the main piece to do here Sub3.x

cell build session file vs netpyne description; hoc file written by cell builder cell templates are usually never touched – get a few PRELIM Data – netpyne currently create instances and then reads those instances

what are the metrics of similarity? – how are we going to compare whether the output is good enough

not currently much randomness in the modeldb models

networks have been very program specific in the past and network construction is embodied in the code – 1st step is to

get rid of code accidents

Aim3.x — CAD vs fully automated

CAD – stick randomizer here universals – number of cells, how many types (small # of types based on morphology with minor variations) modelview currently only gets the lowest of low hanging

same heuristics that are used for cells can be used for networks - whole pool of connections

draw circles around all the types;
in that type are there connection to every other type or to a few types
given a projection is there a constant or a distribution (make some usual assumptions – eg, no 2 projections to 1 cell –

without replacement); what are the statistics of the projections; what those 35 cells are is random ?

try one with 3 types - CA3 – try to catch up with ernie

HBP - read in 1e4 morphologies of 100 morphoelectric sites and Peter’s principal

BRAIN -> modeldb -> we will reverse engineer M1 which we’re already doing

figure of merit – size of the original code vs size of the final netpyne description

how to create the network – how to grow networks; growing networks; find some emergent rules

just use GA on a set of parameters – cell params and weights learning network – RL or other, homeostatic plasticity grow from axon growth cone rules – or GA on probabilities 2 step growning – grow the learning rules and then grow the network emergent properties – dynamical properties, properties of network organization (not an inverse problem) do the netpyne solutions give networks with dynamical properties that are similar to that seen with the grown network

a lot of times algorithms for analysis need to look at places that you have the answer – you don’t care that it’s not

exact; need to see that reduction problem from a 3D net to a stylized Bush and Sejnowsi – reduce to a small cell that captures the dynamics of the big cell (Pinsky and Rinzel, Michele, Ted, HBP)

approach of having a test example of algorithms – explore algorithms used for optimization - does your fitness function

of getting back an optimal similar to the other model – Sean Carver did this Bayesian optimization of ion channels

develop a model (HH type channel model); 2. output data; 3. learn model with Bayesian algorithm from single channel data

1000 Na channels so a noisy set – what experiments can you do on this system to get back the parameters of the channel Bayesian == maximum likelihood – a parameterized model ; a noisy system; let’s say only param is gbar – let’s retrieve it orig data had randomness in it – any model could create this but likelihood is low if the params isn’t correct but even if correct won’t get the data exactly since seed wasn’t same practical difficulty markov model contained some hidden parameters so that had some states for which had no data one problem with least squares on trajectories is that least squares depends on the ‘fact’ that every point on the curve is independent ; it’s a curve not a bunch of random points – least squares is missing the parameterization

two kinds of noise – process noise and experimental noise (artifact)

forcing it to be on the original data trajectory – what’s the likelihood that it’s actually on that trajectory – extremely small so work with log(likelihood)

doing an AP in a model – you are close to threshold so any kid of noise will cause a spike at any time or no spike at all

statistical philosophy – you can take any amount of data out of my distribution; want to know the mean of my distro

can take different ones and end up with a lot of means – is your distribution of means which should be gaussian? what is distro of the mean – this now is recursive mean can be very different from what you find expectation value if the anomalous card had a very large value associated with it maximum likelihood - 95% likelihood but now can run 1e6 times and see ideally would see that 95% but only see 90% so is 90% close enough to 95% so that that is the same follow

new way of organizing grants – this is what we published in last 5 years – is this good enough to continue

LEMD == low-entropy model description, simplified model description, high-level model description, logical model description

Aim 1 RxD part is innovative

subcellular connectivity – dendritic connectivity; deal with regions of the cell have to define the regions – make a description that is indep of the specific morphology NeuronDB canonical cells – medial proximal distal; basal apical axonal; 1st order, 2nd order, 3rd order; yfrac, distances from soma linear dec in IH, IA are functions of distance from soma do we want to find basilar dendrites? – define cells

hoc uses strings and python traditionally uses callables but we’re using strings here; lambdas are not particularly interesting

give an explain as to why we use strings rather than lambda func – nothing exists until it’s instantiated

why do we use python – python gives flexibility for things like error checking interactively

arbitrary keys (with arbitrary values) can be used in values

Aim 4: GUI enter rules into a form that computes the netpyne file

fill in the blanks pictorial stuff - drag&drop? problems of the outsider – what is the meaning of the terms that are obvious to the inventor? – eg probability defined by a sin function?? incorporate – pdf – probability distribution function biased distribution since no cells with too many or too few connections – use convergence or divergence in netpyne

how to do parallel setup – integrate rendezvous rank in netpyne so all the implicit loops are parallel loops

split up aims based on User Experience: Aim 1 – include GUI?)

vs Aim 2: BackEnd – parallel setup (rendezvous rank, use hash to find the node where you will rendezvous), evolutionary algorithms, (5,3,7).__hash__()%1024 # 87 so can rendezvous on 14 – #40 wants that and #375 has it so can find out where to send (send the key of key:value or send the hash) collecting data can also use this use %10 to collect 10 files in 10 places

instantiating on 64k nodes – everyone has the description but no one can see the whole model

can get in TACC or SDSC

running tests and benchmarks – standard tests

outer loop of targets (exist on a particular node), inner loop of sources

does everyone need to know about all targets – we hope not gid contains no information about the cell unless you ask for it

can hash the dict for describing a pop2pop connection – node has a target or set of targets that know what set of sources

they are looking for – the source population doesn’t exist anywhere (it’s just a conceptual object; NB the source population (or target pop) will not fit on any node) – src gids have to declare themselves and made note of their node location NO - not so easy since the rules at the target will be intersections and unions of rules known at a source UHOH being when src->targ based on distance and then only define based on relation of src to targ pc.gidconnect(srcgid) for cell in targets: for netpyne_sources ,,,, relationship of target to source – no common hash – each one has a hash – the target knows what sources it’s looking for and the sources know what targets it’s looking for

distance – can be handled in general - the hash is the rule – everyone now in the same place – the master

handle with a specialized algorithm; any rule that involves pruning is ok; as long as only get a reasonable size subet

testing on megamachines is one reason we want to have a bunch of models to work with – Traub, CA3, M1, Soltesz

Aim 2 could be parallel assistance, fun with parallel

rendezvous rank for setup
load balance – can always reduce to a round robin
parameter optimization evolve,GA (subworlds)

subworlds: doing allgather needs to be restricted to a subset of ranks so that those are doing the same problem

then use bulletin board to throw out the networks – bulletin board is a master/worker style of computation could do bulletin board as a ‘distributed bulletin board’ (could be wasteful) – Linda idea was distributed BB extremes is 1 subworld or 1 node per network and all bulletin board

what are the modern ideas for distributed databases

model, view, controller

graphical introspection – what created this graph? – do something by analogy; find something similar to what you want to do and modify

declarative programming vs procedural – Backus notation

http://maude.cs.uiuc.edu/maude2-manual/html/maude-manualch1.html is there a grammar of networks? can we invent one? would it be useful? – NO, NO, NO the more that things correspond to reality the more likely they are to persist whereas when you are abstract a slight change in perspective obviates the prior results

Nengo – draw lines between networks

representation of the network instance – exhaustive — each cell has the full set of properties repeated or reference to

some shared prototype – each of the cells which are closely related have a shared representation with separate randomizers

appropriate scaling – typically scale down the weights if increase the convergence with increased network size

once you get up to 1e4 no longer allow more convergence with increased network size scale is critical for debugging since need to work with a small networks or hybrid networks reduce to ART_CELLS — can do that

generate the network and save the network – save to disk, save on subworlds, modify a model in netpyne

remove a cell and dump the network back to netpyne – idea of cell death different from another indstance of the network

16dec02 meeting at yale

This tool will subsequently be expanded to allow it to automate the process of employing new data being gathered by ongoing efforts (BRAIN, HBP, Allen) to define specific brain networks and their cellular constituents.

– eveything data-driven, clarify to what extent

NEURON, including the intracellular molecular reaction chains for metabotropic synapses, a level of detail not provided by most neural simulators. We are also building in the ability to readily handle variant hybrid networks, where some cells are modeled in great detail (MC cells)

–

All of the parts, all of this complexity, will remain under the umbrella of a single conceptual model designed using NetPyNE

–

NetPyNE already supports the use of a simple spike-timing dependent plasticity (STDP) mechanism, which has been used to develop a model of sensorimotor cortex that learns to drive a musculoskeletal arm model via reinforcement learning.
provide facilities for readily varying hybrid networks so that, for example, the many runs needed for parameter exploration would be done with only a few MC cells,

– still complicated to switch in a biological sense – need to take into account if connections change as a func of more biophys cell model – computer-assisted

SA1 - comparison of original Traub model with netpyne version
Dont have space to put everyhting or say already done it – appropriate prelim results that give credibility to what you are going to do:

– eg. Traub model

Netpyne is an organizational method for promoting conceptual control

– the more compact, lower entropy – tends toward entropy, but also organizes into conceptual blocks without having the whole model in your head

without programming expertise - red flags

– use templates, can change parameters in isolated ways

high-level parameters to describe complex synaptic distribution

– include figs of sCRACM -> syn distribution

code optimization methods might be required to make the process feasible

– algorithmic optimization

A second premise is that there is not enough memory in any one process to hold a mapofwhichranksholdwhichobjects.

– a map supplied by single process does not scale

example of within dependent pre-and post-synaptic cell condition – = not dist dependent
The simple random distribution of cells evenly on all processors provides a performance baseline for comparison with more sophisticated distribution algorithms, such as dynamic load balance and least processing time algorithm

– only used round-robin or least-processing time

However, it transiently requires twice the memory of the model, does not handle (the rare case) of mod files that use POINTER, and typically shows no improvement for single thread models

– not true, significant

gid compression is only available if each processor owns fewer than 256 cells

16 – from 12 to 2 – could be to 3 or 4
the NEURON Subworlds feature has not been extensively used or tested and is based on an old implementation. Its efficiency

should be evaluated and the implementation improved if deemed necessary. bulletin board does nto scale because of master rank –

remove grid search
reverse eng - spy skulduggery – remove

Questions

GPU will be available for all users of just HBP? include in grant that netpyne will support?

yes will be available in core neuron - so open source, can be added to netpyne

load balancing techniques? papers

Netpyne integrated with NEURON?

(moved to Github issues - 16may31)

- unify cellmodel

cell prop conditions:

– cellType=RS; cellModel=HH – cellType=RS; cellModel=Izhi

provide option to insert distribute mechanism or point process
single cell class
then add support for 2007a - spikes+record from izh._ref_V
INCLUDE BOTH HH AND IZHI IN SAME RULE AND USE CELLMODEL TAG IN CELL TO CHOOSE WHICH TO USE!!

- Overwrite files optional (timestamp)

- Default simConfig

- Import cells with generic set of arguments

- Use mechanismType to make importCell more robust

https://www.neuron.yale.edu/neuron/static/new_doc/modelspec/programmatic/mechtype.html https://www.neuron.yale.edu/neuron/static/new_doc/programming/mechstan.html http://neurosimlab.org/ramcd/pyhelp/mechstan.html#MechanismStandard - 2nd example

use params with initial underscore ‘_var’ to indicate used by netpyne, not mechanism/pointprocess variable - add to docu!

Check Izhi2007a

Distinguish syns from other point processes

no way to do it
assume ‘syn’ in name – establish naming convention – added to doc
syn: true, false, false (.is_netcon_target() = True; .has_net_event(i) = False; .is_artificial() = false)
artcell: true, true?, true
izh2007a: true, true, false
izhi2007b: true, false, false

MechanismType.is_netcon_target() Syntax: boolean = mt.is_netcon_target(i) Description: The i’th point process has a NET_RECEIVE block and can therefore be a target for a NetCon object.

MechanismType.has_net_event() Syntax: boolean = mt.has_net_event(i) Description: The i’th point process has a net_event call in its NET_RECEIVE block and can therefore be a source for a NetCon object. This means it is NetCon stimulator or that the point process can be used as an artificial neural network cell.

MechanismType.is_artificial() Syntax: boolean = mt.is_artificial(i) Description: The i’th point process is an ARTIFICIAL_CELL and can therefore be a source for a NetCon object. This means it is NetCon stimulator or that the point process can be used as an artificial neural network cell.

This seems to have, but does not, equivalent functionality to has_net_event() and was introduced because ARTIFICIAL_CELL objects are no longer located in sections. Some ARTIFICIAL_CELLs such as the PatternStim do not make use of net_event in their implementation, and some PointProcesses do use net_event and must be located in sections for their proper function, e.g. reciprocal synapses.

Make list of data saved part of simCfg

simConfig, netParams, net, simData
added to docu

Test importing Friesen, mainen, traub cell models

Read python names in importCell

– use dir; check if section; compare with secList

‘synReceptor’ check for syn type not name

Associate_gid from axon

section[‘spikeGenLoc’] = 0.5

HDF5

tried h5py (http://docs.h5py.org/en/latest/quick.html)

Saving output as example-20160108_200403.hdf5… Traceback (most recent call last): File “init.py”, line 24, in <module> f.sim.saveData() # save params, cell info and sim output to file (pickle,mat,txt,etc) File “/usr/site/nrniv/pypkg/netpyne/sim.py”, line 346, in saveData hickle.dump(dataSave, f.cfg[‘filename’]+’.hdf5’, mode=’w’) File “/u/salvadord/anaconda/lib/python2.7/site-packages/hickle.py”, line 376, in dump dumper(obj, h5f, **kwargs) File “/u/salvadord/anaconda/lib/python2.7/site-packages/hickle.py”, line 300, in dump_dict _dump_dict(obj, hgroup, **kwargs) File “/u/salvadord/anaconda/lib/python2.7/site-packages/hickle.py”, line 282, in _dump_dict _dump_dict(dd[key], new_group, **kwargs) File “/u/salvadord/anaconda/lib/python2.7/site-packages/hickle.py”, line 269, in _dump_dict hgroup.create_dataset(“%s” % key, data=dd[key], **kwargs) File “/u/salvadord/anaconda/lib/python2.7/site-packages/hickle.py”, line 88, in create_dataset return super(H5GroupWrapper, self).create_dataset(*args, **kwargs) File “/u/salvadord/anaconda/lib/python2.7/site-packages/h5py/_hl/group.py”, line 94, in create_dataset dsid = dataset.make_new_dset(self, shape, dtype, data, **kwds) File “/u/salvadord/anaconda/lib/python2.7/site-packages/h5py/_hl/dataset.py”, line 79, in make_new_dset tid = h5t.py_create(dtype, logical=1) File “h5t.pyx”, line 1389, in h5py.h5t.py_create (h5py/h5t.c:13046) File “h5t.pyx”, line 1463, in h5py.h5t.py_create (h5py/h5t.c:12893) TypeError: Object dtype dtype(‘O’) has no native HDF5 equivalent

tried hdf5storage (http://pythonhosted.org/hdf5storage/introduction.html#getting-started)

Traceback (most recent call last): File “init.py”, line 24, in <module> f.sim.saveData() # save params, cell info and sim output to file (pickle,mat,txt,etc) File “/usr/site/nrniv/pypkg/netpyne/sim.py”, line 348, in saveData hdf5storage.write(dataSave, filename=f.cfg[‘filename’]+’.hdf5’) File “build/bdist.macosx-10.5-x86_64/egg/hdf5storage/__init__.py”, line 1399, in write

File “build/bdist.macosx-10.5-x86_64/egg/hdf5storage/__init__.py”, line 1318, in writes

File “build/bdist.macosx-10.5-x86_64/egg/hdf5storage/lowlevel.py”, line 114, in write_data File “build/bdist.macosx-10.5-x86_64/egg/hdf5storage/Marshallers.py”, line 1557, in write NotImplementedError: Dictionaries with non-unicode keys are not supported: ‘netParams’

try this: http://stackoverflow.com/questions/16705274/fastest-way-to-convert-a-dicts-keys-values-from-str-to-unicode
converted to unicode using dict2utf8() func, but now get segmentation fault for all except simConfig dict
empty dicts {} caused seg fault – replaced with []

Enable providing num spikes in Netstim (paper)

Set v_init before calling init() (paper)

netParam param
‘vinit’ param in section
automatically set vinit=vr for izhis - not needed cause fixed mod
add to doc

Conv and div conn (paper)

made conn very flexible by allowing function for: prob, conv, div, weight and delay
funcs can include spatial/distance variables; and others defined in netParams
can implement a large range of functionalities with very generic implementation

Add x,z,xnorm,znorm positions (paper)

Plot cell from specific population (paper)

and or one of each population

Allow to define xrange and zrange of pops

also xfracrange and zfracrange
for fixed cell density, easy, just don’t calcualte rand
for functional density more complex
make generic so no distinction between coordinates

Fix NetStim connectivity

no convergent or divergent; only full and prob

fix so lambda funcs saved as string

setup.py pip install (paper)

timing (paper)

Record subset of time, and only record traces for subset of cells

recordTraces = [‘all’, ‘P’, 1]
plotTraces = [‘all’, ‘P’, 1]
pops means just 1 cell of that pop!
automatically add elements of plotTraces to recordTraces!!

Fix pops starting with same characters

EMstim == EM !!

Add support for SectionLists (paper)

also when importing

Rename point process ‘_type’ -> ‘mod’, ‘_loc’ -> ‘loc’

Import synapses into synMechParams

Cell position in ellipsoid

Subcellular synaptic connectivity (paper)

distribution of syns across segments based on empirical map
concept of logical connection subsumes subset of unitary connections

simple LFP (paper)

simple voltage sum

Plot conn matrix (paper)

by cells and populations

import/export NeuroML (paper)

padraig (see notes above)

Functions to load simConfig, netParams, net, simData (paper)

importCell can modify h variables (eg. celsius) (paper)

maybe just note in documentation that shouldnt modify h.vars - undesired effects

import python lists of sections (equivalent hoc sectionLists)

Plasticity (paper?)

Hebbian
STDP

Stimulation - iclamp,vclamp, opto, microelectrode (paper?)

maybe iclamp for paper?
can implement via pointp in cell properties? - no, so can assign to multiple
create new struct stimParams

netParams[‘stimParams’] = [] netParams[‘stimParams’][‘sourceList’].append({‘label’: ‘Input_1’, ‘type’: ‘IClamp’, ‘delay’: 0.1, ‘dur’: 0.5, ‘amp’: 0.5}) netParams[‘stimParams’][‘stimList’].append({ ‘source’: ‘Input_1’, ‘sec’:’soma’, ‘loc’: 0.5, ‘conditions’: {‘pop’:’PYR’, ‘cellList’: [0,5]}})

in NeuroML:

Upload examples to netpyne repo (paper)

Evol params batch
virtual arm with RL+STDP
robot control

User input error checking

eg. getattr(a,b, DEFAULT VALUE!)

Travis CI

test automatically examples

Set positions in Neuron’s 3D space

LFPy

cliff?

Save plots to file

Conn/props just in master node

normalized transfer entropy (nTE)

cliff?

Granger causality

cliff?

3D animation of network

use johns code

Add support for point neuron outside of section

eg. IntFire1,2,3,4 ARTIFICIAL_CELL means 3 : that this model not only has a NET_RECEIVE block and does NOT 4 : have a BREAKPOINT but is not associated with a 5 : section location or numerical integrator. i.e it does not 6 : refer to v or any ions or have a POINTER. It is entirely isolated 7 : and depends on discrete events from the outside to affect it and 8 : affects the outside only by sending discrete events.

outside of section so need different method - maybe not worth?

RxD

/u/samn/m1dyst -> models/m1dyst https://www.neuron.yale.edu/neuron/static/docs/rxd/index.html

recreating spikes

Setting up the re-simulation utilizes {\tt PatternStim} to feed the spikes back into the single cell. The setup code is

pattern = h.PatternStim() pattern.fake_output = 1 pattern.play(tvec, idvec)

Saving at intervals

modify instantiated network

don’t destroy NEURON objects, so can interact with them
load net and provide new set of params

Add sim options for efficient parallel sim

add timing of computational vs exchange
load balance
cvode.cache_efficient
compress spikes
bin queue

Save space using slots

http://tech.oyster.com/save-ram-with-python-slots/

Make simulator-indep using NeuroML internal format

GUI

Test adaptive exp cell (no Neuron code)

student doing

Make opensource with contributions from community

There are several ways to contribute to PyNN:

reporting bugs, errors and other mistakes in the code or documentation; making suggestions for improvements; fixing bugs and other mistakes; adding or maintaining a simulator backend; major refactoring to improve performance, reduce code complexity, or both.

Replace search in list with next()

a = next((i for i in userInput if i in wordsTask), None)

Netstim with different types of random input

emphasize alpha, beta, gamma
external pre-recorded signal

Allow range of properties

eg. ‘gnabar’:[0.11, 0.13] and each cell being assigned a random value within that range

Wrappers to change properties interactively both of NEURON+netpyne object

maybe can define generic wrappers for some of the most common cases; maybe include within same global wrapper that checks which needs to be used

generate interactive net struct from code

collapse/expand

http://neurosimlab.org/netpyne/_images/netstruct.png

add units

uS for synMech weight

Netstim default start

Plot cell subset of each population, indicating relative index

? Connection function with specific connections

fromList
relativePreIds
relativePostIds
connList

Update Cm of section based on C of Izhi

Add option to save different data (specs, instance, output) to different file names

including dat (txt) format

Files

sdnotes.org

Latest commit

History

sdnotes.org

File metadata and controls

15may27 Modularized framework for NEURON/Python network simulations with MPI

Overview

Bill’s description of what this repo should be

README with description of the different files

Known issues, potential improvements, and alternative implementations (outdated - see TODO list at end of file)

- Saving only happens at the end (instead of at fixed intervals)

- Need to add a DEBUG mode or similar (like mpiHHTut had)

- Check structure of data saved to file

- Some attributes/variables of the class ‘Cell’ and ‘Pop’ are not used but still initialized to [ ]

- Labels (eg. AMPA, IT, HH,…) are defined by assigning numeric values to variables

- stimuli.py not tested with this model

- analysis.py still has many functions which havent been adapted to new structure

- Some variables inherited from cliff still have too long names, eg. backgroundSpikevec

- Lines too long, need to split using \

- if increase p.backgroundNoise > 0.0 get memory error

- Find a way to save lambda funcs (can’t be pickled)

- Clearing vector in simdata now gives error because also has python variables

- ‘Cell’ class record method uses eval() which is unsafe

- currently setup to record traces from all cells

- bug when saving cell traces using multiple cores

- Load/save net+sim parameters from file

- Store net and sim params in dictionaries inside params.py

- Use dict of tags/attributes for pop params and for cells

Define conn rules based on pair of tag/value for pre and post

- Add $Id$ hg info

- Rename main.py with init.py

Remove popType and connType variables - infer from dict keys

Remove popid from popParams - can use list index

Replace topClass and subClass with projection-type, type (neurotransmitter involved?)

email to Ben

Synapse

15dec28 Convert into python package

16jan26 Possibly making more simulator-indep using NeuroML-based format

NeuroML format

Issues

cells

connections

synapses

population

pulse generator (netstim?)

chat with Bill

chat with Padraig

16jan27 Consolidate connectivity functions

Full conn -> fullConn

combine convConn, divConn, and randConn?

Random conn with N conns per postsyn cell (convergence) -> convConn

Random conn with N conns per presyn cell (divergence) -> divConn

Random conn with fixed, function (including distance-dep) prob, weight, delay -> probConn

fixed

func

func dist-dep

implementation

lambda with kwargs - NO

lambda with args - NO!

string

efficiency

errors

comparison to PyNN

class AllToAllConnector(allow_self_connections=True, safe=True, callback=None)[source]

class OneToOneConnector(safe=True, callback=None)[source]

class FixedProbabilityConnector(p_connect, allow_self_connections=True, rng=None, safe=True, callback=None)[source]

class FromListConnector(conn_list, column_names=None, safe=True, callback=None)[source]

class FromFileConnector(file, distributed=False, safe=True, callback=None)[source]

class ArrayConnector(array, safe=True, callback=None)[source]

class FixedNumberPreConnector(n, allow_self_connections=True, with_replacement=False, rng=None, safe=True, callback=None)

class DistanceDependentProbabilityConnector(d_expression, allow_self_connections=True, rng=None, safe=True, callback=None)[source]

class IndexBasedProbabilityConnector(index_expression, allow_self_connections=True, rng=None, safe=True, callback=None)[sour

class SmallWorldConnector(degree, rewiring, allow_self_connections=True, n_connections=None, rng=None, safe=True, callback=N

class CSAConnector(cset, safe=True, callback=None)[source]

comparison to NEST

ConvergentConnect

DivergentConnect

RandomConvergentConnect

RandomDivergentConnect