Newsgroups: comp.ai.neural-nets,comp.soft-sys.matlab
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!europa.eng.gtefsd.com!news.mathworks.com!yeshua.marcam.com!usc!math.ohio-state.edu!caen!msuinfo!uchinews!ellis!gal2
From: gal2@ellis.uchicago.edu (Jacob Galley)
Subject: Drive-reinforcement NN, in Matlab?
Message-ID: <1994Nov9.061547.27295@midway.uchicago.edu>
Sender: news@uchinews.uchicago.edu (News System)
Reply-To: gal2@midway.uchicago.edu
Organization: University of Chicago
Date: Wed, 9 Nov 1994 06:15:47 GMT
Lines: 24
Xref: glinda.oz.cs.cmu.edu comp.ai.neural-nets:19969 comp.soft-sys.matlab:9374


Hi, I am currently implementing Klopf's drive-reinforcement learning
paradigm within Matlab, using the Neural Network toolbox.  Klopf's
model is a feed-forward network that uses a complex variant of Hebb's
learning rule.  Hebb said that if the activation level of two
connected nodes at time t were high, then increase the strength of the
connection.  Instead of just measuring the level of activity at t,
Klopf's model measures the CHANGE in the activity level of each node
over time.  If the activity of each is increasing, then increase the
strength of the connection.  Because of this, I have to keep track of
the previous n states of the network.

Has anyone out there tried to implement drive-reinforcement?

In Matlab?

Perhaps we can compare our techniques and results.

Jake.

-- 
"What we are really trying to do is to install the sort of junta that America 
is trying to get rid of in Haiti, and then run like hell."
					   <-- an anonymous UN aide in Somalia
