Donald F. Beal

A generalised quiescence search algorithm

Abstract This paper describes how the concept of a null move may be used to define a generalised quiescence search applicable to any minimax problem. Experimental results in the domain of chess tactics show major gains in cost effectiveness over full-width searches, and it is suggested that null-move quiescence may be almost as widely useful as the alpha-beta mechanism. The essence of the mechanism is that null moves give rise to bounds on position values which are more reliable than evaluations. When opposing bounds touch, they create a single value which is more reliable than ordinary evaluations, and the search is terminated at that point. These terminations are prior to any alpha-beta cutoffs, and can lead to self-terminating searches.

Heuristic Programming in Artificial Intelligence 3

The first Soviet Computer Olympiad 2nd Computer Olympiad reports Go intellect wins two gold medals databases in Awari an architecture for a sophisticated mechanical bridge player design and implementation of a Chinese chess program reviving the game of checkers exploratory learning in the game of Go investigating Tsumego problems with RisiKo applying retrograde analysis to Nine Mens Morris experiments with the NegaC search - an alternative for Othello endgame search deep forks in strategic maps brute force in search of games of imperfect information artificial intelligence or stochastic relaxation - simulated annealing challenge the smart game board as a tool for game progammers.

First Results from Using Temporal Difference Learning in Shogi

This paper describes first results from the application of Temporal Difference learning [1] to shogi. We report on experiments to determine whether sensible values for shogi pieces can be obtained in the same manner as for western chess pieces [2]. The learning is obtained entirely from randomised self-play, without access to any form of expert knowledge. The piece values are used in a simple search program that chooses shogi moves from a shallow lookahead, using pieces values to evaluate the leaves, with a random tie-break at the top level. Temporal difference learning is used to adjust the piece values over the course of a series of games. The method is successful in learning values that perform well in matches against hand-crafted values.

Temporal difference learning applied to game playing and the results of application to shogi

This paper describes the application of temporal difference (TD) learning to minimax searches in general, and presents results from shogi. TD learning is used to adjust the weights for evaluation features over the course of a series of games, starting from arbitrary initial values. For some games, to obtain weights accurate enough for high-performance play will require the TD learning phase to make use of minimax searches. A theoretical description of TD applied to minimax search is given, and we discuss how the theoretical characteristics of the method interact with practical considerations. These include the depth of search appropriate for successful learning and the use of self-play to enable the algorithm to be independent of human knowledge. We then report on experiments that obtained values for use in shogi-playing programs. Unlike chess, shogi has no generally agreed standardized set of values for pieces, so there is more need for machine learning. We compare our machine-learnt values, obtained without any human knowledge input, with hand-crafted values. TD learning was successful in obtaining values that performed well in matches against hand-crafted values.

DESIGN OF A PROCESSOR ELEMENT FOR A HIGH PERFORMANCE MASSIVELY PARALLEL SIMD SYSTEM

This paper describes the architecture of the General Purpose with Floating Point support (GPFP) processing element, which uses the expansion of circuitry from VLSI advances to provide on-chip memory and cost-effective extra functionality. A major goal was to accelerate floating point arithmetic. This was combined with architectural aims of cost-effectiveness, achieving the floating-point capability from general-purpose units, and retaining the 1-bit manipulations available in the earlier generation. With a 50 MHz clock each PE is capable of 2.5 MegaFlops. Normalized to the same clock rate, the GPFP PE exceeds first generation PEs by far, namely the DAP by a factor of 50 and the MPP by a factor of 20, and also outperforms the recent MasPar design by a factor of four. A 32×32 GPFP array is capable of up to 2.5 GigaFlops and 6500 MIPS, on 32-bit additions. These speedups are obtained by architectural features rather than increased width of data-handling and are combined with parsimonious use of circuitry compatible with massively parallel fabrication. The GPFP also incorporates Reconfigurable Local Control (RLC), a technique that combines a considerable degree of local autonomy within PEs and microcode flexibility, giving the machine improved general-purpose programmability in addition to floating-point numerical performance.

Solving Chinese chess endgames by database construction

Abstract This paper describes an improved retrograde algorithm used for work on constructing Chinese chess endgame databases. Significant differences from databases for (western) chess are examined, and their influence on database construction noted. Results from Chinese chess yield surprises for human Chinese chess experts, overturning previous human analysis over many years by top players. The aegp-aaee ending is a theoretical win, and not as previously believed, a draw. We also consider the computational resources required to perform these computations.

Recent progress in understanding minimax search

Lookahead search has been known for a long time to be effective in tackling problems which can be cast in minimax form (mainly game playing, but others include maintaining unstable balance against gravity, and business decisions). Recent results have shown that the benefit of lookahead depends on the structure inherent in the problem, and even that there exist some minimax problems for which lookahead search is DIS-advantageous. This paper reviews those results and then discusses algorithms which can be interpreted as recognising the structure of local areas of the search in order to control search expansion. Such algorithms can be orders of magnitude more cost-effective than search using alpha-beta alone.

GPFP: an array processing element for the next generation of massively parallel supercomputer architectures

No abstract available