Written by Brian Blais
What is Plasticity?
Plasticity is a package of programs, with a convenient interface,
used to run simulations of single cells and networks of neurons. It is used
to explore BCM synaptic modification, Hebbian learning, ICA, and others. It
accompanies the book Theory of Cortical
Plasticity by Leon Cooper, Nathan Intrator, Brian Blais, and Harel
Shouval. It is available from World Scientific.
basic principle of the simulations is the presentation of pattern and noise
vectors to a network of neurons, with specified lateral connectivity and
synaptic modification rule. The pattern vectors are either patches from
images or simply read from a matlab .mat file as columns of a matrix. The
noise vectors are generated during run-time. The interface lets you specify
many things including
the dimensionality of the inputs
number of channels (left/right eye, ON/OFF channels, etc.)
the pattern input files
generation of input files from a specified correlation function
the noise mean, variad> the noise mean, variance and type (uniform, gaussian, etc.)
the lateral connectivity
the learning rule
constraints on the weights (normalization, saturation limits, etc.)
number of neurons
number of iterations
what information to display
Plasticity is written in Matlab, using MEX
functions for speed. The package will not work without these binary
MEX functions. The current package includes binaries for Microsoft Windows,
Linux, and Mac OSX. It should compile on anything that has
Matlab. If you do compile it on another architecture, please send an email and it will be
included in future revisions.
Matlab 6 or greater (it may
work in Matlab 5, but since I do not have access to Matlab 5 anymore I can't
say for sure).
(possibly) a C compiler
(last updated Mar 16 2004)
Tar File (for all platforms)
Warning for WinZip users: When uncompressing this file, turn
off the option to do "Smart CR/LF Conversion" on TAR files. This
messes up the .mat files, and causes Matlab to crash rather
Guide to Plasticity. This
documentation is pretty sparse, but will get you started.
These files are about 15 Meg compressed, 25 Meg expanded.
Just download the files, unpack them in a directory (usually within your
matlab directory), and run 'plasticity'.
There is a little bit of documentation here.
Brian Blais' research page, with many
of online publications