Results 11  20
of
96
Transposition Table Driven Work Scheduling in Distributed Search
 IN 16TH NATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI'99
, 1999
"... This paper introduces a new scheduling algorithm for parallel singleagent search, transposition table driven work scheduling, that places the transposition table at the heart of the parallel work scheduling. The scheme results in less synchronization overhead, less processor idle time, and less ..."
Abstract

Cited by 15 (2 self)
 Add to MetaCart
This paper introduces a new scheduling algorithm for parallel singleagent search, transposition table driven work scheduling, that places the transposition table at the heart of the parallel work scheduling. The scheme results in less synchronization overhead, less processor idle time, and less redundant search effort. Measurements on a 128processor parallel machine show that the scheme achieves nearlyoptimal performance and scales well. The algorithm performs a factor of 2.0 to 13.7 times better than traditional workstealingbased schemes.
A Performance Analysis of TranspositionTableDriven Scheduling in Distributed Search
 IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
, 2002
"... This paper discusses a new workscheduling algorithm for parallel search of singleagent state spaces, called TranspositionTableDriven Work Scheduling, that places the transposition table at the heart of the parallel work scheduling. The scheme results in less synchronization overhead, less proce ..."
Abstract

Cited by 13 (8 self)
 Add to MetaCart
(Show Context)
This paper discusses a new workscheduling algorithm for parallel search of singleagent state spaces, called TranspositionTableDriven Work Scheduling, that places the transposition table at the heart of the parallel work scheduling. The scheme results in less synchronization overhead, less processor idle time, and less redundant search effort. Measurements on a 128processor parallel machine show that the scheme achieves closetolinear speedups; for large problems the speedups are even superlinear due to better memory usage. On the same machine, the algorithm is 1.6 to 12.9 times faster than traditional workstealingbased schemes.
A solution to the GHI problem for depthfirst proofnumber search. Manuscript in preparation
 In 7th Joint Conference on Information Sciences
, 2003
"... Abstract The GraphHistory Interaction (GHI) problem is a notorious problem that causes gameplaying programs to occasionally return incorrect solutions. This paper presents a practical method to cure the GHI problem for the case of the dfpn algorithm. Results in the game of Go with the situational ..."
Abstract

Cited by 12 (3 self)
 Add to MetaCart
Abstract The GraphHistory Interaction (GHI) problem is a notorious problem that causes gameplaying programs to occasionally return incorrect solutions. This paper presents a practical method to cure the GHI problem for the case of the dfpn algorithm. Results in the game of Go with the situational superko rule show that the overhead incurred by our method is very small, while correctness is always guaranteed. Keywords: GHI problem, dfpn algorithm, Kawano's simulation 1 Introduction 1.1 Motivation Developing high performance gameplaying programs has been the subject of AI research for over 50 years. Gameplaying programs typically employ lookahead search to improve their move decisions. Efficient search algorithms improve the strength of their programs. For example, Thompson showed that there is a strong positive correlation between the explored depth of the search tree and the strength of a chessplaying system [12]. Therefore, programmers have invested a large amount of resources to enhance their search engines.
Computer Chess And Search
 ARTICLE PREPARED FOR THE 2ND EDITION OF THE ENCYCLOPEDIA OF ARTIFICIAL INTELLIGENCE, S. SHAPIRO (EDITOR), TO BE PUBLISHED BY JOHN WILEY, 1992.
, 1991
"... ..."
Replacement Schemes for Transposition Tables
 ICCA Journal
"... Almost every chess program makes use of a transposition table, typically implemented as a large hash table. Even though this table is usually made as large as possible, subject to memory constraints, collisions occur. Then a choice has to be made which position to retain or to replace in the table, ..."
Abstract

Cited by 9 (3 self)
 Add to MetaCart
Almost every chess program makes use of a transposition table, typically implemented as a large hash table. Even though this table is usually made as large as possible, subject to memory constraints, collisions occur. Then a choice has to be made which position to retain or to replace in the table, using some replacement scheme. This article compares the performance of seven replacement schemes, as a function of transpositiontable size, on some chess middlegame positions. A twolevel table, using the number of nodes in the subtree searched as the deciding criterion, performed best and is provisionally recommended. 1 Introduction Chess programs analyze positions while building trees. However, a closer look shows that the search space could better be explored by graphs, due to the fact that a position can be reached by several orderings of moves. Such resultant positions are known as transpositions. When encountering a position again, the size of the search tree can be reduced conside...
Search in Trees with Chance Nodes
 MASTER'S THESIS
, 2004
"... While much of the research done in heuristic search has concentrated on deterministic domains, not much work has been done to investigate search techniques in stochastic domains other than statistical sampling methods. When full search is required, Expectimax is often the algorithm of choice. Howeve ..."
Abstract

Cited by 8 (1 self)
 Add to MetaCart
While much of the research done in heuristic search has concentrated on deterministic domains, not much work has been done to investigate search techniques in stochastic domains other than statistical sampling methods. When full search is required, Expectimax is often the algorithm of choice. However, Expectimax is a fullwidth search algorithm. A class of algorithms called *Minimax were developed by Bruce Ballard to improve on Expectimax's runtime. They allow for cutoffs in trees with chance nodes similar to how AlphaBeta allows for cutoffs in Minimax trees. This thesis presents new performance results for Expectimax, as well as Star1 and Star2 (the two main *Minimax algorithms), in realworld domains. Ballard's work is verified and new insights into move ordering and probe successor selection are presented.
Symmetry detection in general game playing
 In Proceedings of the IJCAI09 Workshop on General Game Playing (GIGA’09
, 2009
"... We develop a method for detecting symmetries in arbitrary games and exploiting these symmetries when using tree search to play the game. Games in the General Game Playing domain are given as a set of logic based rules defining legal moves, their effects and goals of the players. The presented metho ..."
Abstract

Cited by 8 (1 self)
 Add to MetaCart
We develop a method for detecting symmetries in arbitrary games and exploiting these symmetries when using tree search to play the game. Games in the General Game Playing domain are given as a set of logic based rules defining legal moves, their effects and goals of the players. The presented method transforms the rules of a game into a vertexlabeled graph such that automorphisms of the graph correspond with symmetries of the game. The algorithm detects many kinds of symmetries that often occur in games, e.g., rotation and reflection symmetries of boards, interchangeable objects, and symmetric roles. A transposition table is used to efficiently exploit the symmetries in many games. 1
A scalable machine learning approach to go
 in Advances in Neural Information Processing Systems 19
, 2007
"... Go is an ancient board game that poses unique opportunities and challenges for AI and machine learning. Here we develop a machine learning approach to Go, and related board games, focusing primarily on the problem of learning a good evaluation function in a scalable way. Scalability is essential at ..."
Abstract

Cited by 7 (0 self)
 Add to MetaCart
(Show Context)
Go is an ancient board game that poses unique opportunities and challenges for AI and machine learning. Here we develop a machine learning approach to Go, and related board games, focusing primarily on the problem of learning a good evaluation function in a scalable way. Scalability is essential at multiple levels, from the library of local tactical patterns, to the integration of patterns across the board, to the size of the board itself. The system we propose is capable of automatically learning the propensity of local patterns from a library of games. Propensity and other local tactical information are fed into a recursive neural network, derived from a Bayesian network architecture. The network integrates local information across the board and produces local outputs that represent local territory ownership probabilities. The aggregation of these probabilities provides an effective strategic evaluation function that is an estimate of the expected area at the end (or at other stages) of the game. Local area targets for training can be derived from datasets of human games. A system trained using only 9 × 9 amateur game data performs surprisingly well on a test set derived from 19 × 19 professional game data. Possible directions for further improvements are briefly discussed. 1