self-organization software, papers, web demos

[James A. Bednar]

The following software package for self-organization in laterally
connected maps is now available from the UTCS Neural Networks Research
Group website http://www.cs.utexas.edu/users/nn. The software has been
developed in the RF-LISSOM project of modeling the primary visual
cortex, and is intended to serve as a starting point for computational
studies of development and function of perceptual maps in general.

Abstracts of two recent papers on the RF-LISSOM project, on
segmentation and on internal pattern generators, are also included
below. Other papers and demos of the RF-LISSOM software are available
at http://www.cs.utexas.edu/users/nn/pages/research/selforg.html.

- Jim, Yoonsuck, and Risto


Software:
-----------------------------------------------------------------------
RFLISSOM: LATERALLY CONNECTED SELF-ORGANIZING MAPS
http://www.cs.utexas.edu/users/nn/pages/software/abstracts.html#lissom

James A. Bednar and Joseph Sirosh

The LISSOM package contains the ANSI C source code and examples for
training and testing RF-LISSOM.  This implementation supports almost
every modern single-processor workstation, as well as the Cray T3E
massively parallel supercomputer.  It is designed to have full
functionality even when run in batch mode or remote mode, using a
simple but powerful command file format.  It also has an interactive
command-line prompt accepting the same commands, which makes it easy
to test and explore different options in real-time.  Because of the
focus on batch/remote use, it does not have a GUI, but it does create
a wide variety of images for analysis and testing.  It can display
these images immediately or save them for later viewing.  Sample
command files are provided for running orientation and
ocular-dominance map simulations on a variety of network and machine
sizes.  Extensive documentation is also included, all of which is also
available via online help where appropriate.

This implementation currently supports the LISSOM algorithm only,
but it can serve as a good starting point for writing a batch-mode
neural-network or related simulator.  In particular, it includes
independent and general-purpose routines for interprocessor
communication, platform independence, console messaging, error
handling, command-file/command-prompt processing, parameter
setting/ retrieving/ bounds-checking, expression parsing, and online
help.



Papers:
-----------------------------------------------------------------------
PATTERN-GENERATOR-DRIVEN DEVELOPMENT IN SELF-ORGANIZING MODELS

James A. Bednar and Risto Miikkulainen
In James M. Bower, editor, Computational Neuroscience: Trends
in Research, 1998, 317-323.  New York: Plenum, 1998. (7 pages)

http://www.cs.utexas.edu/users/nn/pages/publications/abstracts.html#bednar.cns97.ps.Z

Self-organizing models develop realistic cortical structures when given
approximations of the visual environment as input. Recently it has been
proposed that internally generated input patterns, such as those found
in the developing retina and in PGO waves during REM sleep, may have the
same effect.  Internal pattern generators would constitute an efficient
way to specify, develop, and maintain functionally appropriate
perceptual organization.  They may help express complex structures from
minimal genetic information, and retain this genetic structure within a
highly plastic system.  Simulations with the RF-LISSOM orientation map
model indicate that such preorganization is possible, providing a
computational framework for examining how genetic influences interact
with visual experience.

The results from this paper can be reproduced with the LISSOM
software described above.  

-----------------------------------------------------------------------
SELF-ORGANIZATION AND SEGMENTATION IN A LATERALLY CONNECTED
ORIENTATION MAP OF SPIKING NEURONS

Yoonsuck Choe and Risto Miikkulainen
Neurocomputing, in press (20 pages)

http://www.cs.utexas.edu/users/nn/pages/publications/abstracts.html#choe.neurocomputing98.ps.Z

The RF-SLISSOM model integrates two separate lines of research on
computational modeling of the visual cortex.  Laterally connected
self-organizing maps have been used to model how afferent structures
such as orientation columns and patterned lateral connections can
simultaneously self-organize through input-driven Hebbian
adaptation. Spiking neurons with leaky integrator synapses have been
used to model image segmentation and binding by synchronization and
desynchronization of neuronal group activity. Although these approaches
differ in how they model the neuron and what they explain, they share
the same overall layout of a laterally connected two-dimensional
network.  This paper shows how both self-organization and segmentation
can be achieved in such an integrated network, thus presenting a unified
model of development and functional dynamics in the primary visual
cortex.

A demo of RF-SLISSOM can be run remotely over the internet at
http://www.cs.utexas.edu/users/nn/pages/research/selforg.html.