A Theory of Heuristic Information in Game-Tree Search
General Material Designation
[Book]
First Statement of Responsibility
by Chun-Hung Tzeng.
.PUBLICATION, DISTRIBUTION, ETC
Place of Publication, Distribution, etc.
Berlin, Heidelberg
Name of Publisher, Distributor, etc.
Springer Berlin Heidelberg
Date of Publication, Distribution, etc.
1988
PHYSICAL DESCRIPTION
Specific Material Designation and Extent of Item
(x, 107 pages 22 illustrations)
SERIES
Series Title
Symbolic computation., Artificial intelligence.
CONTENTS NOTE
Text of Note
Introduction --; Games and Minimax Values --; Heuristic Game-Tree Searches --; Probability Spaces and Martingales --; Probabilistic Game Models and Game Values --; Heuristic Information --; Estimation and Decision Making --; Independence and Product-Propagation Rules --; Estimation of Minimax Values in Pb-Game Models --; Estimation of Minimax Values in Gd-Game Models --; Conclusions --; Appendix --; References --; Subject Index.
SUMMARY OR ABSTRACT
Text of Note
This book presents the use of imperfect information (called heuristic information) in game-tree search. Its purpose is to investigate the theoretical background of the use of heuristic information in game-tree search. Computer programs playing games usually search the game-tree to a reasonable depth with a static evaluation function and make decisions based upon backed-up values. Since the information in either the backed-up values or the values returned directly by the static evaluation function is often imperfect, decision making is usually not optimal. Also, pathological cases show why intuition about game-tree search is not always correct. This book introduces a mathematical formulation of heuristic information and a theoretical model of game-tree search. In this model, notions of game-tree search are formulated in mathematical terms and a sound mathematical theory of heuristic information is developed. The conventional pathological cases disappear in this theory. This book is accessible to the general AI community, for example first year graduate students who have completed an introductory AI course and have at least some background in probability. The book is also a foundation for further work on game-tree search as well as on heuristic information in general AI.