-
Notifications
You must be signed in to change notification settings - Fork 0
/
MgII_NMF_HP.py
486 lines (378 loc) · 18.2 KB
/
MgII_NMF_HP.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
#Constants------------------------------------
first_line_wave = 2796.3543
second_line_wave = 2803.5315
rf_line_sep = second_line_wave-first_line_wave
#Doublet Finder Hyperparameters
rf_err_margain = 1.0
kernel_smooth = 2
med_filt_size_tight = 19
med_filt_size_wide = 39
med_filt_size_fit = 85
snr_threshold_low = 1.5
snr_threshold_high = 2.5
#MCMC Hyperparamets
sub_region_size = 40
min_accept_frac = 0.45
min_tol = 100
nwalkers = 32
ndim = 5 #not really a hyper
chain_length = 15000
chain_discard = 1000
#----------------------------------------------------
#multiprocessing seteup
import multiprocessing
nprocesses = multiprocessing.cpu_count() // 8
#standard imports
import numpy as np
import os
import h5py
import sys
import desispec.io
import fitsio
from astropy import modeling
#from astropy.table import Table
import emcee
from desimodel.footprint import radec2pix
from desispec.coaddition import coadd_cameras
#doublet finder setup
from astropy.convolution import Gaussian1DKernel
from astropy.convolution import convolve
from scipy.signal import medfilt
kernel = Gaussian1DKernel(stddev = kernel_smooth)
from operator import itemgetter
from itertools import *
fitter = modeling.fitting.LevMarLSQFitter()
model = modeling.models.Gaussian1D()
#using the main and sv cmx
from desitarget.cmx.cmx_targetmask import cmx_mask
from desitarget.targets import main_cmx_or_sv
import NMF_quasar_SEDs_decomposition_module as NMF
NMF_basis = NMF.load_quasar_NMF_basis()
def MgII_Model(theta,x):
z,a1,a2,s1,s2 = theta
#determine peak centers
m1 = (z+1)*first_line_wave
m2 = (z+1)*second_line_wave
#Generate Model
model = a1*np.exp((-(x-m1)**2)/(2*s1**2))+a2*np.exp((-(x-m2)**2)/(2*s2**2))
return model
#basic chi2 function
def chi2(obs,cal,var,reduced = True):
c2 = np.sum((obs-cal)**2 / var)
if reduced:
return c2/len(obs)
else:
return c2
#likelihood fnc
def log_likelihood(theta, x, y, yerr):
#generative model
model = MgII_Model(theta,x)
#error into variance
sigma2 = yerr ** 2
#Actual Likelihood fnc
return -0.5 * np.sum((y - model) ** 2 / sigma2 + np.log(sigma2))
#prior fnc, could contain more info on reasonable redshifts, heights and widths
def log_prior(theta,z_low,z_high):
z,a1,a2,s1,s2 = theta
#if -100 < a1 < 100 and -100 < a2 < 100 and 0 < s1 and 0 < s2 and z_low < z < z_high:
if 0 < s1 and 0 < s2 and z_low < z < z_high:
return 0.0
return -np.inf
#probability fnc
def log_probability(theta, x, y, yerr, z_low, z_high):
lp = log_prior(theta,z_low,z_high)
if not np.isfinite(lp):
return -np.inf
return lp + log_likelihood(theta, x, y, yerr)
def Doublet_MCMC(Doublet,x_spc,y_flx,y_err):
#spectrum info
x_spc,y_flx,y_err = x_spc,y_flx,y_err
#determine redshift and appropriate line_sep
z = float(Doublet[0])
line_sep = rf_line_sep*(1+z)
peak = int(Doublet[1])
#set sub region values, bounded by [0,x_spc-1]
srh = min(len(x_spc)-1,peak+sub_region_size)
srl = max(0,peak-sub_region_size)
#determine max and min z in window (or lowest/highest values possible if at edge of wavelength space)
z_low = x_spc[srl]/first_line_wave-1
z_high = x_spc[srh]/first_line_wave-1
#define subregion in x and y
reg_wave = x_spc[srl:srh]
reg_flx = y_flx[srl:srh]
reg_err = y_err[srl:srh]
init_Amp1 = -float(Doublet[4])
init_Amp2 = -float(Doublet[6])
init_StdDev1 = float(Doublet[5])
init_StdDev2 = float(Doublet[7])
#define initial theta (TODO: Reconsider guesses for m,b)
initial = [z,init_Amp1,init_Amp2,init_StdDev1,init_StdDev2]
#could widen this inital guess range, don't think it matters though
p0 = initial + 1e-4 * np.random.randn(nwalkers, ndim)
#setup sampler with args
sampler = emcee.EnsembleSampler(nwalkers, ndim, log_probability, args = [reg_wave,reg_flx,reg_err,z_low,z_high])
#burn-in
state = sampler.run_mcmc(p0, 100)
sampler.reset()
#MCMC can hit errors during running, so wrap in try/except
try:
state = sampler.run_mcmc(state,chain_length)
except ValueError:
return
#extract estimate autocorrelation time, number of times chain is longer than autocorr time and mean_accept_frac
tau = sampler.get_autocorr_time(discard = chain_discard,quiet = True)
implied_tol = chain_length/max(tau)
mean_accept_frac = np.mean(sampler.acceptance_fraction)
#mcmc quality cuts, if passed return the samples
if(implied_tol > min_tol and mean_accept_frac > min_accept_frac):
#extract MCMC info
flat_samples = sampler.get_chain(flat=True,discard=chain_discard)
return (flat_samples)
else:
return
#Takes spectra info and returns list of doublets that pass snr criteria
def doublet_finder(continuum,x_spc,y_flx,y_err,min_z):
#calculate residual
residual = continuum-y_flx
#find indices where residuals are positive
pos_inds = np.where(residual > 0)[0]
#from: https://stackoverflow.com/questions/7352684/how-to-find-the-groups-of-consecutive-elements-in-a-numpy-array, credit user: unutbu
cons_groups = np.array(np.split(pos_inds, np.where(np.diff(pos_inds) != 1)[0]+1),dtype=object)
#find groups longer than 2 entries
minidx_mask = np.array([arr.size > 2 for arr in cons_groups])
#check groups of 3 of more indices for snr
absorb_lines = []
for group in cons_groups[minidx_mask]:
#calculate snr
snr = np.sum(residual[group])/np.sqrt(np.sum(y_err[group]**2))
#set central index to be the highest value
cen = np.nanargmax(residual[group])
#skip entries with implied redshifts below min_Z
#could run this earlier but here seems fine
if x_spc[group][cen]/first_line_wave - 1 < min_z:
continue
if snr > snr_threshold_low:
try:
model = modeling.models.Gaussian1D(amplitude = np.nanmax(residual[group]), mean = x_spc[group][cen], stddev = 0.4)
fm = fitter(model, x = x_spc[group], y = residual[group])
except:
print(model,group)
#save the line parameters as well as group of indices, snr and central index
absorb_lines.append([fm.parameters[1],group,snr,fm.parameters[0],fm.parameters[2],cen])
#convert to np array
absorb_lines = np.array(absorb_lines,dtype=object)
#calculate what implied redshift would be if each feature was first/second line of MgII doublet
firstline_z = absorb_lines[:,0]/first_line_wave - 1
secondline_z = absorb_lines[:,0]/second_line_wave -1
#calculate the allowed redshift difference implied by the restframe seperation, restframe error margain and firstline redshift
z_diff = ((1+firstline_z)*(first_line_wave+rf_line_sep+rf_err_margain))/second_line_wave - 1 - firstline_z
#comparing each firstline to it's corresponding second line (note slicing) we can check if below redshift difference
doublet_loc = np.where(secondline_z[1:]-firstline_z[:-1] < z_diff[:-1])[0]
#save a doublet for each case where redshift difference is acceptable and snr thresholds are met
doublets = []
for loc in doublet_loc:
line1 = absorb_lines[loc]
line2 = absorb_lines[loc+1]
#check if leading line (2796) meets high snr threshold
if line1[2] > snr_threshold_high:
#append the redshift, center index, snrs and initial line amplitudes and widths
doublets.append([firstline_z[loc],line1[5],line1[2],line2[2],line1[3],line1[4],line2[3],line2[4]])
return np.array(doublets)
def Doublet_Detection(cat_subset,hp):
hp_str = str(hp)
if int(hp) < 100:
hp_short = '0'
else:
hp_short = hp_str[0:-2]
surveys = np.unique(cat_subset['SURVEY'])
#format output directory and filename
out_dir = '{}/{}/{}'.format(out_base_dir,hp_short,hp_str)
print(out_dir)
out_fn = 'MgII-Abs-Chains-{}.hdf5'.format(hp_str)
#names of model params for h5py formatting
names = ['z','Amp1','Amp2','StdDev1','StdDev2']
#if output file path does not exist create it
if not os.path.exists(out_dir):
os.makedirs(out_dir)
#full format out filepath
out_file = os.path.join(out_dir,out_fn)
#open h5py buffer to write out file
with h5py.File(out_file, "w") as f:
for survey in surveys:
#format coadd directory and filename
in_dir = os.path.join(reduction_base_dir, "healpix",survey,"dark",hp_short,hp_str)
in_fn = "coadd-{}-{}-{}.fits".format(survey, "dark", hp)
#read in specobj file, and also coadd
specfile = os.path.join(in_dir, in_fn)
#print(specfile)
specobj = desispec.io.read_spectra(specfile)
coadd_specobj = coadd_cameras(specobj)
#grab wavelength grid as this will be same for all spectra
x_spc = coadd_specobj.wave["brz"]
cat_srv_sub = cat_subset[cat_subset['SURVEY']==survey]
#find indices of specobj that have a targetid value in the catalog subset
#i.e. find the indices of the QSO_cat entries in the specobj
#QSO_inds = np.where(np.isin(specobj.target_ids(),cat_subset['TARGETID'])==True)[0]
#iterate over qso indices and the related shift from catalog subset
#little opaque here but clean
for entry in cat_srv_sub:
redshift = entry['Z']
TARGETID = entry['TARGETID']
#have to pull out from inside listed list
ind = np.where(specobj.target_ids() == TARGETID)[0][0]
#calculate wavelength of MgII abs at Lya emission (we don't search this region)
min_z = (1216*(1+redshift)/first_line_wave)-1
#grab flux and error date for specific spectra
y_flx = coadd_specobj.flux["brz"][ind]
y_err = 1/np.sqrt(coadd_specobj.ivar["brz"][ind])
#apply gaussian smoothing kernel
smooth_yflx = convolve(y_flx,kernel)
#estimate continuum using combination of two median filters
c_tight = medfilt(y_flx,med_filt_size_tight)
c_wide = medfilt(y_flx,med_filt_size_wide)
#length of filter scale for normalizing
f_scale = len(c_tight)
#weight tight filter higher at low z and wide filter higher at high z
doublet_cont = c_tight*(1-np.arange(f_scale )/f_scale) + c_wide*(np.arange(f_scale )/f_scale)
#run doublet finder
doublets = doublet_finder(doublet_cont,x_spc,smooth_yflx,y_err,min_z)
#NEW code for NMF/medfilt continuum conditions
NMF_fail = False
Cont_method = ''
#using NMF estimator, will sometimes fail so t/e wrapped. If it fails we want to try medfilt continuum so set bool
try:
out = NMF.NMF_normalization_v2(np.log10(x_spc),y_flx,coadd_specobj.ivar["brz"][ind],redshift,NMF_basis)
except:
NMF_fail = True
#if NMF failed, always use medfilt continuum
if NMF_fail:
#calculate new medilt and set as fitting_cont, set cont_method
medfilt_cont = medfilt(y_flx,med_filt_size_fit)
fitting_cont = medfilt_cont
Cont_method = 'Medfilt'
#if NMF is poorly fit
elif out[-1] > 4.0:
#calculate new medilt and chi2
medfilt_cont = medfilt(y_flx,med_filt_size_fit)
medfilt_chi2 = chi2(y_flx,medfilt_cont,coadd_specobj.ivar["brz"][ind])
#choose continuum with better chi2
if medfilt_chi2 < out[-1]:
fitting_cont = medfilt_cont
Cont_method = 'Medfilt'
else:
fitting_cont = out[-2]
Cont_method = 'NMF'
else:
fitting_cont = out[-2]
Cont_method = 'NMF'
for doublet in doublets:
chain = Doublet_MCMC(doublet,x_spc,y_flx-fitting_cont,y_err)
#if the chain returned is actually an object (i.e. MCMC didn't crash)
if not chain is None:
#write hp name for entry, with MgII Abs redshift and QSO em redshift
z_MgII = np.round(np.percentile(chain[:, 0],50),decimals = 5)
z_QSO = np.round(redshift,decimals=5)
out_str = '{}_{}_{}_{}_{}_{}_{}_{}'.format(TARGETID,survey,hp,z_MgII,z_QSO,doublet[2],doublet[3],Cont_method)
#create h5py group, TODO: way to check if out_str is an h5py group w/o try/except?
try:
grp = f.create_group(out_str)
except ValueError as e:
#if group already exists then this MgII feature is already recorded, move to next entry
continue
for k in range(ndim):
grp.create_dataset(names[k],dtype = float,data = chain[:,k])
#These pased as command line args
redux = str(sys.argv[1])
reduction_base_dir = '/global/cfs/cdirs/desi/spectro/redux/{}/'.format(redux)
#open catalog
if redux == 'fuji' or redux == 'guadalupe':
out_base_dir = '/pscratch/sd/l/lucasnap/MgII_Abs_Chains/{}-NMF'.format(redux)
QSOcat_fp = '/global/cfs/cdirs/desi/users/edmondc/QSO_catalog/{}/QSO_cat_{}_healpix_only_qso_targets.fits'.format(redux,redux)
QSOcat = fitsio.read(QSOcat_fp,'QSO_CAT')
#restrict to only dark time exposures
QSOcat = QSOcat[QSOcat['PROGRAM'] == 'dark']
elif redux == 'iron':
out_base_dir = '/pscratch/sd/l/lucasnap/MgII_Abs_Chains/{}-NMF'.format(redux)
QSOcat_fp = '/global/cfs/cdirs/desi/survey/catalogs/Y1/QSO/{}/QSO_cat_{}_main_dark_healpix_only_qso_targets_vtest.fits'.format(redux,redux)
QSOcat = fitsio.read(QSOcat_fp,'QSO_CAT')
#restrict to only dark time exposures
QSOcat = QSOcat[QSOcat['PROGRAM'] == 'dark']
elif redux == 'fuji_all':
redux = 'fuji'
reduction_base_dir = '/global/cfs/cdirs/desi/spectro/redux/{}/'.format(redux)
out_base_dir = '/pscratch/sd/l/lucasnap/MgII_Abs_Chains/{}-ALL'.format(redux)
QSOcat_all_fp = '/global/cfs/cdirs/desi/users/edmondc/QSO_catalog/{}/QSO_cat_{}_healpix.fits'.format(redux,redux)
QSOcat_all = fitsio.read(QSOcat_all_fp,'QSO_CAT')
#restrict to only dark time exposures
QSOcat_all = QSOcat_all[QSOcat_all['PROGRAM'] == 'dark']
QSOcat_qso_fp = '/global/cfs/cdirs/desi/users/edmondc/QSO_catalog/{}/QSO_cat_{}_healpix_only_qso_targets.fits'.format(redux,redux)
QSOcat_qso = fitsio.read(QSOcat_qso_fp,'QSO_CAT')
QSOcat_qso = QSOcat_qso[QSOcat_qso['PROGRAM'] == 'dark']
print('Initial all catalog length: {}'.format(len(QSOcat_all)))
print('Initial qso catalog length: {}'.format(len(QSOcat_qso)))
#match_TID,all_idx,qso_idx = np.intersect1d(QSOcat_all['TARGETID'],QSOcat_qso['TARGETID'],return_indices=True)
all_both_mask = np.isin(QSOcat_all['TARGETID'],QSOcat_qso['TARGETID'])
QSOcat = QSOcat_all[~all_both_mask]
print('After removing qso target entries {} entries remain in catalog'.format(len(QSOcat)))
#create list of unique healpix values
hp_vals = radec2pix(64,QSOcat['TARGET_RA'],QSOcat['TARGET_DEC'])
hp_unique = np.unique(hp_vals)
#routine to check if hp values have been completed
#make a mask of healpix directories that have/have not been succesfully complete
hp_complete_mask = []
for hp_val in hp_unique:
#recast as str
hp_str = str(hp_val)
if int(hp_val) < 100:
hp_short = 0
else:
hp_short = hp_str[0:-2]
#format filepath to MgII output
out_dir = '{}/{}/{}'.format(out_base_dir,hp_short,hp_str)
h5_fn = 'MgII-Abs-Chains-{}.hdf5'.format(hp_str)
h5_fp = os.path.join(out_dir,h5_fn)
#see if a valid h5 file is written into path (TODO: a way to do this without try/except?)
try:
f = h5py.File(h5_fp,'r')
#if it is (i.e. no error thrown) we have completed this hp_val (mask[val]=True)
hp_complete_mask.append(True)
#if a valid h5 file hasn't been written. we can check if it exists and remove it, then proceed. Either way (mask[val]=False)
except OSError as e:
if(str(e)[0:19]=='Unable to open file' and os.path.exists(h5_fp)):
os.remove(h5_fp)
hp_complete_mask.append(False)
#np-ify (TODO: better way to do this?)
hp_complete_mask = np.asarray(hp_complete_mask)
#find subset of incomplete hp_dirs
hp_incomplete = hp_unique[~hp_complete_mask]
print(len(hp_unique))
print(len(hp_incomplete))
#take percentage of hp_incomplete corresponding to passed argv
#currently defeaults to 10 percent interval
if(len(sys.argv)==4):
hp_incomplete = hp_incomplete[int(len(hp_incomplete)*int(sys.argv[3])/100):int(len(hp_incomplete)*(int(sys.argv[3])/100+0.25))]
print(len(hp_incomplete))
print('Running MgII Absorption Finder on healpix directories: {}.'.format(hp_incomplete))
#If there are no incomplete directories, survey is complete. Exit
if(len(hp_incomplete)==0):
print('All healpix directories are already completed for release: {}'.format(redux))
sys.exit()
#Function pool will pass each hp_unique entry to
def RunFinder_HPind(hp_val):
#pull subset of QSO_cat in healpix area
cat_sub = QSOcat[hp_vals==hp_val]
#run feature detection and MCMC step
Doublet_Detection(cat_sub,hp_val)
print('Done: ',hp_val)
if (str(sys.argv[2])=='safe'):
#run a single random entry with a single process
nprocesses = 1
print('{} processes being utilized'.format(nprocesses))
#RunFinder_HPind(np.random.choice(hp_incomplete))
RunFinder_HPind(hp_incomplete[0])
if (str(sys.argv[2])=='full'):
#Setup and run pool
pool = multiprocessing.Pool(processes = nprocesses)
print('{} processes being utilized'.format(nprocesses))
#run pool
pool.map(RunFinder_HPind, hp_incomplete)