Learning to score final positions in the game of Go

Authors:
Erik C. D. van der Werf;H. Jaap van den Herik;Jos W. H. M. Uiterwijk
Affiliations:
Department of Computer Science, Institute for Knowledge and Agent Technology, Universiteit Maastricht, MD Maastricht, The Netherlands and GN ReSound Research & Developement, Algorithm R&D, Horsten ...;Department of Computer Science, Institute for Knowledge and Agent Technology, Universiteit Maastricht, MD Maastricht, The Netherlands;Department of Computer Science, Institute for Knowledge and Agent Technology, Universiteit Maastricht, MD Maastricht, The Netherlands
Venue:
Theoretical Computer Science - Advances in computer games
Year:
2005

Citing 10
Cited 0

Temporal difference learning and TD-Gammon

Communications of the ACM
GO Is Polynomial-Space Hard

Journal of the ACM (JACM)
Statistical Pattern Recognition: A Review

IEEE Transactions on Pattern Analysis and Machine Intelligence
Static analysis of life and death in the game of Go

Information Sciences—Informatics and Computer Science: An International Journal
Computer Go: an AI oriented survey

Artificial Intelligence
Computer Go

Artificial Intelligence - Chips challenging champions: games, computers and Artificial Intelligence
Honte, a go-playing program using neural nets

Machines that learn to play games
The Golem Go Program

The Golem Go Program
Learning to predict life and death from Go game records

Information Sciences: an International Journal
Learning to estimate potential territory in the game of go

CG'04 Proceedings of the 4th international conference on Computers and Games

Quantified Score

Hi-index	0.01

Visualization

Abstract

This article investigates the application of machine-learning techniques for the task of scoring final positions in the game of Go. Neural network classifiers are trained to classify life and death from labelled 9 × 9 game records. The performance is compared to standard classifiers from statistical pattern recognition. A recursive framework for classification is used to improve performance iteratively. Using a maximum of four iterations our cascaded scoring architecture (CSA*) scores 98.9% of the positions correctly. Nearly all incorrectly scored positions are recognised (they can be corrected by a human operator). By providing reliable score information CSA* opens the large source of Go knowledge implicitly available in human game records for automatic extraction. It thus paves the way for a successful application of machine learning in Go.