[computer-go] Learning approach to go

[Eric Baum <ebaum@xxxxxxxxxxx>]

Hi Nicol.

At http://www.whatisthought.com/eric.html you can find 
powerpoint slides describing the learning approach 
Igor Durdanovic and I took to producing a Go program. 
We tried to cut subgraphs out of a big game database, 
after reducing the game position to a topological graph,
value them, store a big database of valued subgraphs, 
and evaluate new positions out of overlapping pieces. 
We went so far as to build a cluster to do it.
It didn't ultimately work because we couldn't find
a semantically meaningful enough way of cutting out 
the subgraphs.

My other papers from my NEC days are up there also, 
including Bayesian game playing, Hayek, evolutionary 
computing, GA, and DNA computing.

BTW, this has little in common with how *What is 
Thought?* models human approach -- this was more 
of a computer science approach, an attempt to use 
the strengths of the computer to attack Go analogous 
to (but different from) how you can use the strengths 
of the computer to attack chess.

Sorry for the awkwardness in posting-- its hard from a digest.
I'm switching over.

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://computer-go.org/mailman/listinfo/computer-go