rich@GTE-LABS.CSNET (Rich Sutton) (03/06/86)
Self-Organization, Memorization,
and Associative Recall of Sensory Information
by Brain-Like Adaptive Networks
Tuevo Kohonen, Helsinki University of Technology
The main purpose of thinking is to forecast phenomena that take place
in the environment. To this end, humans and animals must refer to a
complicated knowledge base which is somewhat vaguely called memory.
One has to realize the two main problem areas in a discussion of memory:
(1) the memory mechanism itself, and (2) the internal representations of
sensory information in the brain networks.
Most of the experimental and theoretical works have concentrated on the
first problem. Although it has been extremely difficult to detect memory
traces experimentally, the storage mechanism is theoretically still the
easier part of the problem. Contrary to this, it has been almost a
mystery how a physical system can automatically extract various kinds
of abstraction from the huge number of vague sensor signals. This paper
now contains some novel views and results about the formation of such
internal representations in idealized neural networks, and their
memorization. It seems that both of the above functions, viz. formation
of the internal representations and their storage, can be implemented
simultaneously by an adaptive, self-organizing neural structure which
consists of a great number of neural units arranged into a
two-dimensional network. A number of computer simulations are presented
to illustrate both the self-organized formation of sensory feature maps,
as well as associative recall of activity patterns from the distributed
memory.
When: March 14, 1:00 pm
Where: GTE Labs 3-131
Contact: Rich Sutton, Rich@GTE-Labs.CSNet, (617)466-4133