Efficient Neural Net alpha-beta-Evaluators (1994)
| Venue: | Proceedings of the International Conference on Artificial Neural Networks |
BibTeX
@INPROCEEDINGS{Heinz94efficientneural,
author = {Alois P. Heinz},
title = {Efficient Neural Net alpha-beta-Evaluators},
booktitle = {Proceedings of the International Conference on Artificial Neural Networks},
year = {1994},
pages = {635--638},
publisher = {Springer-Verlag}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Introduction We describe a new artificial neural network for the conceptual realization of ff-fi- evaluators which are extensively used as the central part of many game playing programs. Then we present a very efficient implementation of this network that provides a speed-up of recall time in the order of magnitudes on conventional computers as compared to sequential implementations of traditional neural network architectures. The ff-fi-evaluator is a function of three arguments, a game position p and the boundaries ff and fi of a real interval [ff; fi], also called the ff-fi-window. It returns a real value v(p; ff; fi) 2 [ff; fi] that may be considered as a measure of the real strength of the position adjusted to the nearest point of the<
Keyphrases
efficient neural net alpha-beta-evaluators efficient implementation real strength conventional computer real value game position central part conceptual realization many game recall time sequential implementation traditional neural network architecture ff-fi evaluator new artificial neural network real interval ff
