July 19, 2000 

Evolved Neural Network Successfully Defends Checkers Challenge at 2000 Congress on EC

La Jolla, CA - Dr. David Fogel, executive vice president and chief scientist of Natural Selection, Inc.^Ò , reports that the neural network that was created from over 800 generations of an evolutionary algorithm remains undefeated after 25 challenge matches with people registered at the 2000 Congress on Evolutionary Computation. "Honestly, I was hoping to be able to give the $100 prize to someone, but I suppose I'll have to save if for another competition at a future meeting," said Dr. Fogel. The best-evolved neural network only received one significant challenge during the four-day event, from Professor Edmund Burke of the University of Nottingham in the United Kingdom. "My collaborator, Kumar Chellapilla, and I call the neural network 'Anaconda' because it often seems to win by constricting its opponent's mobility. I saw that a few times in the course of the 25 matches that I played here," offered Dr. Fogel. He continued, "I want to thank all the participants in the challenge. It's been a lot of fun and serves to demonstrate that evolutionary algorithms can create solutions to difficult problems even when relying on very little information." Fogel and Chellapilla's evolving neural networks were only given the type, location, and number of pieces on the board. Everything else about how to interpret the position of the checkers had to be evolved during the course of play. Anaconda has demonstrated itself to be on par with human experts, despite having no preprogrammed "human expertise." "We think this is an important demonstration of the capability of an evolutionary approach to problem solving," said Chellapilla, a senior staff scientist at Natural Selection, Inc.® "It provides a clear example of what we can expect from evolutionary algorithms, and when you think that we only used a Pentium II 400MHz machine for evolving the neural networks, there's no doubt that evolutionary algorithms can be applied to solve real-world practical problems."

Natural Selection, Inc.^Ò 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 diagnosis of breast cancer from film-screen mammograms, optical character recognition, and a variety of military and industrial projects.

For more information, please contact:

Dr. David B. Fogel,
Executive Vice President & Chief Scientist,
Natural Selection, Inc.,
3333 N. Torrey Pines Ct.,
Suite 200,
La Jolla, CA, 92037
tel: (619) 455-6449
fax: (619) 455-1560
dfogel@natural-selection.com

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