Site Map Webmaster
All materials © Natural Selection, Inc.®, 2003, 2004. All rights reserved.
December 21, 2004

Natural Selection, Inc. Publishes New Results on Self-Evolving Chess Program

La Jolla, CA – Natural Selection, Inc. is pleased to announce the publication of new results on research into having a computer teach itself how to play chess. The publication appears in the current issue (December, 2004) of the Proceedings of the IEEE, the flagship journal of the world's largest professional organization of engineers. The paper, authored by Dr. David Fogel, Tim Hays, Sarah Hahn, and James Quon (a nationally ranked chess master), all with Natural Selection, Inc., can be viewed by clicking here.

The research, funded in part by a Phase II Small Business Innovative Research (SBIR) grant from the National Science Foundation*, utilizes a combination of evolutionary computing and artificial neural networks to optimize strategies for playing chess. The computer simulates an evolutionary process that competes different strategies, represented in part by neural networks. Those strategies that win more games and lose fewer games are favored in the evolutionary simulation to become parents for future offspring strategies. The offspring are random variations from their parents. After a series of 50 generations in 10 separate trials, the best-evolved strategy was tested against Pocket Fritz 2.0, a commercial chess program. The results from a series of 12 games indicated that the best-evolved strategy was able to defeat Pocket Fritz 2.0 nine times, with two losses and one draw, earning a so-called performance rating of 2550, on par with grandmasters. All of the moves of each of the 12 games are annotated and can be found by clicking here.

The best-evolved chess program is called Blondie25, and follows earlier research by Dr. David Fogel (CEO of Natural Selection, Inc.) and Kumar Chellapilla on the self-learning checkers program called Blondie24. That program is described in the book Blondie24: Playing at the Edge of AI and is incorporated in the product Evolutionary Checkers starring Blondie24, published by Digenetics, Inc. The new evolved chess strategy is now incorporated in Digenetics, Inc.'s second product, Chess with an Attitude!. Digenetics, Inc. is a sister company of Natural Selection, Inc.

"We're extremely pleased to see the results of our research published in such a prestigious journal as the Proceedings of the IEEE," said Dr. Fogel. "We've already completed research involving a much longer series of generations and are currently testing the performance of the best-evolved chess player. The results so far have been very encouraging. The implications of this research, however, go far beyond chess, for the results demonstrate a consistent ability for a computer to teach itself highly effective strategies in complex situations using an evolutionary process. Such situations are not limited to chess, but extend also to military combat, data mining in homeland security applications, as well as having computer programs control characters in video games," Dr. Fogel explained. "In addition, the use of so-called object neural networks, which focus on specific objects in a scene or specific facets of data, appears to accelerate the evolutionary process on the problems we've studied."

Natural Selection, Inc. was founded in 1993 to address complex problems in industry, medicine, and defense. The company possesses unique expertise in computational intelligence techniques, including evolutionary computation, neural networks, and fuzzy logic. The corporation's research efforts support the discovery of new pharmaceuticals, the automated detection and discovery of important patterns and processes in bioinformatics and medical informatics, computer-assisted diagnosis, and a variety of military and industrial projects.

* This work was sponsored in part by the National Science Foundation under Grants DMI-0232124 and DMI-0349604. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

For more information, please contact:

Dr. David B. Fogel
Chief Executive Officer
Natural Selection, Inc.
3333 N. Torrey Pines Ct.
Suite 200
La Jolla, CA, 92037
tel: (858) 455-6449
fax: (858) 455-1560
dfogel@natural-selection.com

and please visit our web site at http://www.natural-selection.com/.