memory (and SOMs)

mat mats_trash at hotmail.com
Tue Mar 19 13:02:01 EST 2002


> One way to resolve this issue is to understand the mechanism of how
> long-range connections between maps are being made. Unfortunately
> there arnt much known about these aspects (from what I've read).
> But Ive got some more reading...
> 
> I guess the most direct approach is to start tracing the pathways from
> some identified areas in the cortex, such as V1-V2 and so on. Then one
> can find out if the pathways are fixed or variable. Im not sure how
> difficult that would be.

I was re-reading these posts and it seemed relevant to some stuff I've
read recently by Edelman and Tononi's group.  They used random graph
theory, developed by Erdos amongst others, to create an evolutionary
(more properly developmental) simulation of the neocortex.  What they
got was different patterns of connectivity between a large number of
nodes.  They then tested these patterns in such a way as to mimic
electrophysiological responses, thus identifting connectivity patterns
that may be to some extent isomorphic to that of the cortex. 
Interestingly the best-fitting patterns were those that developed
highly interconnected local groups of neurones with relatively sparse
long-range connections between the local groups (columns?).  You could
be highly critical of their model becuase it doesn't include any
alogithms for synaptic plasticity or other known neural processes, but
on the other hand you could envisage that if you use tests-of-function
of a high enough level, modelling these processes is not required.

I think one of their papers on this is freely downloadable from
www.elsevier.com, in the journal Neural Networks.  Also try the
journal Cerebral Cortex.




More information about the Neur-sci mailing list