To: wdl
Subject: WDL Dis. #5
Bcc: dlall

===== WDL Dis. #5 ============================================
From jebara@media.mit.edu  Tue Feb 15 18:42:54 2000
Date: Tue, 15 Feb 2000 18:42:47 -0500
To: Dr John J Weng <weng@cse.msu.edu>
CC: wdl@cse.msu.edu
Subject: Re: WDL Dis #4

John,

Thanks for the pointers. I understand that some of the latest
results might still be upcoming and I'm eagerly awaiting
further publications and the workshop itself. However,
I also found Prof. Mriganka Sur's earlier paper quite intriguing:

"Rewiring Cortex: The Role of Patterned Activity in
Development and Plasticity of Neocortical Circuits",
M. Sur, A. Angelucci and J. Sharma
Journal of Neurobiology, 41: 33-43, October 1999.

It would be interesting to discuss what specifically can be
done about the way we approach representation and learning
at a computational level to encourage such plasticity in
machines.

It often seems easier and more immediate to manually
pick features, design models and engineer a system. Usually,
finding a generalizable principle, criterion or objective
function that will do automatic feature selection, model
selection, etc. is rather tricky. The plasticity or generality
that comes out of it, however, should eventually pay off in
the long run.

-Tony