Oesterreichisches Forschungsinstitut fuer Artificial Intelligence (OeFAI)
Schottengasse 3, A-1010 Wien
Tel.: +43-1-53361120, Fax: +43-1-5336112-77, Email: sec@ai.univie.ac.at
-------------------------------------------------------------------------
VORTRAG
*******
Dr. Hendrik Blockeel
Department of Computer Science
Katholieke Universiteit Leuven, Belgium
Experiments in the use of machine learning for playing Go
Go is a board game with very simple rules but a complex high-level behaviour.
While brute-force approaches (e.g., an alpha-beta search) work well for many
other board games, such as chess, they do not work well for Go because of the
very large number of moves that can be made. It is therefore crucial to
understand the effect of moves in terms of the board patterns that they help to
create or destroy.
The best current computer programs for playing Go are much less good than good
human players, mainly because they lack insight in such patterns. Such insight
is hard to program, so a program would preferably learn about the important
patterns and their interactions itself.
Inductive logic programming is a machine learning technique that has the
potential to identify important patterns and to find out how to exploit them. It
can be hoped that with such an approach improvements to Go playing programs
become possible.
This talk consists of two parts. In a first part the game of Go will be
presented, as well as inductive logic programming. The second part focuses on
the application of inductive logic programming in Go. This will be illustrated
by a number of experiments that show the usefulness of the approach.
Zeit: Donnerstag, 29. Maerz 2001, 18:30 Uhr pktl.
Ort: Oesterreichisches Forschungsinstitut fuer Artificial Intelligence
Schottengasse 3, 1010 Wien 1.
OESTERREICHISCHES FORSCHUNGSINSTITUT
FUER ARTIFICIAL INTELLIGENCE
o.Univ.-Prof.Dr.Robert Trappl