Evolving Algorithms For Constraint Satisfaction

Stuart Bain

School of Information Technology, Griffith University

Mon Aug 09 13:10:00 NZST 2004 in Room S6, Sc

Abstract

Many different algorithms have been proposed for solving constraint satisfaction problems - specifically problems cast as boolean satisfiability. Different algorithms are necessary as the "No Free Lunch" theories hold that no one algorithm is ideal for all types of problems. We propose a novel method of exploring the space of possible algorithms, by evolving specific algorithms for particular constraint problems using Genetic Programming.