At 01:50 PM 2/11/99 +0100, Anders wrote:
>quite similar to what happens in the Bayesian Confidence Propagation
>Neural Networks I'm working with; there the network tries to minimize
>an "energy" based on the mutual information and entropy of the
>data). I'm not sure it will work out in the end, but it is a fresh way
>of looking at things.
Isn't this close to the classic Hopfield net model derived by analogy with
spin glasses?
Damien Broderick
Received on Thu Feb 11 20:36:35 1999
This archive was generated by hypermail 2.1.8 : Tue Mar 07 2006 - 14:45:30 PST