Automating the strategy selection for parallel heuristic search
R. Craig Varnell · 1998
Many Artificial Intelligence (AI) programs attempt, to apply some form of intelligence to problem solving.Problem solving may include planning the moves of a robot, parsing and interpreting a sentence, or playing the opponent in a game of chess.All of these include some form of search where a set of alternatives are evaluated and the most appropriate alternative is selected that could eventually lead to a solution.The search process often requires a great deal of processor time which can seriously affect the program's performance.For time critical applications the use of the fastest processor often does not produce timely results.Even parallel processing does not necessarily reduce the processing time especially if the wrong approach is taken for a problem.With each problem having a different structure it is impossible to use the same approach for all problems.This research has examined a method for automating the selection of a strategy for parallel search.Machine learning, and the C4.5 machine learning system in particular, is used to choose the best strategy based on a problem's structure.