A pattern classification approach to evaluation function learning

DOI10.1016/0004-3702(88)90076-8MaRDI QIDQ1102763zbMATH OpenOpenAlexFDO

Authors Kai-Fu Lee, Sanjoy Mahajan

Publication date 1988

Published in Artificial Intelligence (Search for Journal in Brave)

Full work available at URL https://figshare.com/articles/journal_contribution/A_pattern_classification_approach_to_evaluation_function_learning/6591158

zbMATH Keywords

Bayesian learningevolution functiongame playingOthellopattern classification

Mathematics Subject Classification ID

Classification and discrimination; cluster analysis (statistical aspects) (62H30) Learning and adaptive systems in artificial intelligence (68T05) Pattern recognition, speech recognition (68T10) Other game-theoretic models (91A40)

Review text

We present a new approach to evaluation function learning using classical pattern classification methods. Unlike other approaches to game playing where ad hoc methods are used to generate the evaluation function, our approach is a disciplined one based on Bayesian learning. This technique can be applied to any domain where a goal can be defined and an evolution function can be applied. Such an approach has several advantages: (1) automatic and optimal combination of the features, or terms of the evaluation function; (2) understanding of inter-feature correlation; (3) capability for recovering from erroneous features; and (4) direct estimation of the probability of winning by the evaluation function. We implemented this algorithm using the game of Othello and it resulted in dramatic improvements over a linear evaluation function that has performed at world championship level.

Recommendations

• scientific article; zbMATH DE number 1759679
• scientific article; zbMATH DE number 1875729
• PAC-Bayesian pattern classification with kernels. Theory, algorithms, and an application to the game Go.
• An acquisition of evaluation function for shogi by learning self-play
• A Bayesian approach to relevance in game playing

Cites work

• A comparison and evaluation of three machine learning procedures as applied to the game of checkers
• A world-championship-level Othello program
• scientific article; zbMATH DE number 3436645 (Why is no real title available?)

Cited in

(14)

• A Bayesian approach to relevance in game playing
• Embedding decision-analytic control in a learning architecture
• Achieving superior generalisation with a high order neural network
• Learning game-specific spatially-oriented heuristics
• Probabilistic heuristic estimates
• Practical issues in temporal difference learning
• scientific article; zbMATH DE number 1759679 (Why is no real title available?)
• scientific article; zbMATH DE number 1875729 (Why is no real title available?)
• scientific article; zbMATH DE number 4187136 (Why is no real title available?)
• Learning for efficient search
• Spatial state-action features for general games
• Automatic acquisition of search guiding heuristics
• Learning to score final positions in the game of Go
• Tuning evaluation functions by maximizing concordance

This page was built for publication: A pattern classification approach to evaluation function learning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1102763)