Dr. Levinson and two graduate students will study methods of augmenting or replacing search with subpattern recognition, associative recall, and machine learning. A general adaptive prediction model has been tested in standard search domains for which analytic and performance results are known. This system will now be extended to a more complex domain of a self-learning, pattern-oriented chess program, to be trained against published games and available deep-searching chess programs. The goal is to develop problem-solving strategies resembling those used by intuitive humans experts.
