In this article we present our chess engine Tempo. One of the major difficulties for this type of program lies in the function for evaluating game positions. This function is composed of a large number of parameters which have to be determined and then adjusted. We propose an alternative which consists in replacing this function by an artificial neuron network (ANN). Without topological knowledge of this complex network, we use the evolutionist methods for its inception, thus enabling us to obtain, among other things, a modular network. Finally, we present our results:
an experimental chess engine developed by Mathieu Autonès, Aryel Beck, Phillippe Camacho, Nicolas Lassabe, Hervé Luga, and François Scharffe, using an artificial neuronal network as evaluation function generated by a genetic algorithm. Tempo applies L-systems as elaborated by Egbert Boers and Herman Kuiper [1], to generate modular neural networks whose size is independent of that of the chromosome and crossover tolerant. A population of L-system construction rules is generated to mark the resulting networks according to their capabilities to learn game position evaluations from real games. Simple aspects were learnt at first. Strongest individuals learnt more complex faetures, from material to square control and pawn structure [2].
Table of Contents
See also
Publications
Abstract
Forum Posts
External Links
References
What links here?
Up one Level