[computer-go] Heuristics Search and Learning

["Robin Kramer" <robin@xxxxxxxx>]

In a
> perfect world, if your position evaluation is good enough, your search
> need only be 1 ply deep.

Any algorithm which can be solved recursively can also be solved
iteratively.  There is a theorm that the evaluation heuristic would be
exponential given an exponential search space, so there is an optimization
problem between evaluation heuristic and search.

There are varieties of weight based learning programs that also translate
the data into a space where the data is linearly seperable.  They are
called support vector machines(SVM).  These training algorithms are
garunteed to converge to the optimum classification based on a seperating
hyperplane which is exactly half the "distance" between the groups.  The
problem is that they are slow to evaluate, because a SVM classifier is the
size of the original data set.

Sincerely,

Robin Kramer


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