stosim --folder <path-to-your-experiment-folder> --run --ttests --plots
StoSim is a light-weight framework for parameterised stochastic simulations. The researcher provides the simulation code (written in the programming language of his/her choice), but StoSim takes over a lot of tedious work from there on. StoSim can
- create all necessary computation jobs (from lists of parameter settings)
- execute these jobs on available CPUs (only your computer, the local network or PBS computation cluster)
- handle stochastic repetition of settings and seeding of randomness (for repeatable experiments)
- analyse the results with graphical plots and T-tests (it is easy to analyse results made with specific parameter settings)
- back up code and results (to be able to go back to important milestones)
You can find tutorials and extensive documentation at http://stosim.nicolashoening.de and example simulations in the "example" folder.
pip install stosim
More details and help are in the documentation.
- You need Python 3.3+, 2.7, or 2.6 if you install the
argparse
module locally. - Distributing jobs needs Unix Screens, so if the
screen
command is not available on your system, you need to install it, e.g. on Debian:sudo apt-get install screen
- To use the in-built plotting, you need gnuplot and epstopdf, e.g. on Debian:
sudo apt-get install gnuplot texlive-extra-utils
- To run T-Tests within StoSim, you need Gnu R installed. E.g. on Debian:
sudo apt-get install r-base
More details and help are in the documentation.
Place an experiment configuration (stosim.conf) and your simulation code in a folder of your choice (see basic example in the examples folder). Call:
stosim
This assumes that you placed a configuration file describing your jobs (called stosim.conf
) in the current folder.
It also assumes you want to run the jobs, and in addition perform T-Tests and make plots (if stosim.conf says how).
So the above command is synonymous to:
stosim --folder . --run --ttests --plots
The results will be put in the data
directory, in your folder
(but if you like the plotting/analysis capabilities of StoSim you might never have to look there).
One more example, where you know your stosim.conf is in a different folder and you only want to run:
stosim --folder <path-to-your-experiment-folder> --run
There are more features, e.g. --more
to add more stochastic runs, --list
to inspect data or --snapshot
to make a snapshots of results at a given time that can be loaded back later. Do check out the tutorials where you can learn more.