ObjSim

ObjSim

This is a c++ based object oriented library for simulating spiking neural networks. It was used for simulations in the following publications:

• "Michler F, Eckhorn R, Wachtler T (2009)" Using Spatiotemporal Correlations to Learn Topographic Maps for Invariant Object Recognition. J Neurophysiol 102:953-64.
• "Michler F, Wachtler T, Eckhorn R (2006)" Adaptive Feedback Inhibition Improves Pattern Discrimination Learning. Lecture Notes in Artificial Intelligence 4087:21-32.

Compiling

Running build_all.sh will compile:

• the simulation library csim (source code is in src)
• simulation main programs:
  • simulations/fm/som02/som02.cpp which was used in Frank Michler et al. (2009)
  • minimal/minimal.cpp is a minimal simulation program demonstrating how to use the library.
• some unit tests in mycxxtests using the cxxtest library

Binaries will be installed in install.

Running via command line

You can run the simulations via:

cd install/bin/som02
./som02

settings_som02.cfg contains default simulation parameters that can be overwritten via command line arguments.

Running via Jupyter noteooks

After compiling, simulations can also be started and evaluated from jupyter notebooks.

• Before you must provide a shell script in env/activate_python_env.sh to setup your python environment. You can use env/activate_python_env_EXAMPLE.sh as an example.
• start_jupyter.sh starts jupyter notebook in subdirectory python.
• The notebook python/ipynb/run_objsim_with_gaussian_input.ipynb can be used to run som02.cpp
• python/ipynb/calc_maps.ipynb demonstrates how to load simulation data and calculate topographic maps.