[computer-go] Re: Topological Invariant?

[Eric Baum <ebaum@xxxxxxxxxxx>]

> Does someone know any existing work about topological invariant
 > representation of GO board without loss of game-playing
 > information?

I don't know if you subscribe to  genetic_programming@xxxxxxxxxxxxxxxxx,
this came up in my answer to a question there a few days ago about
some foils I'd posted.

In the foils on A Learning Go Program at
http://www.whatisthought.com/eric.html , I describe a hashing
method that we used to hash subgraphs of what we called the
topological graph, so that we could find if these subgraphs were in
a stored database of subgraphs. The topological graph was 
formed from the Go position simply by shrinking strings to single 
nodes. This of course loses game playing information such as:
the number of stones in a string, which can be vital if the string
is captured to know whether the capturing group can form eyes;
and also who played last in a ko. The solution we found for this problem
was to hash an annotated graph, with appropriate annotations
for the lost information included.
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://computer-go.org/mailman/listinfo/computer-go