A study of mutational robustness as the product of evolutionary computation

This paper investigates the ability of a tournament selection based genetic algorithm to find mutationally robust solutions to a simple combinatorial optimization problem. Two distinct algorithms (a stochastic hill climber and a tournament selection based GA) were used to search for optimal walks in several variants of the self avoiding walk problem. The robustness of the solutions obtained by the algorithms were compared, both with each other and with solutions obtained by a random sampling of the optimal solution space. The solutions found by the GA were, for most of the problem variants, significantly more robust than those found by either the hill climbing algorithm or random sampling. The solutions found by the hill climbing algorithm were often significantly less robust than those obtained through random sampling. Categories and Subject Descriptors I.2.8 [Artificial Intelligence]: Problem Solving, Control Methods, and Search General Terms Algorithms Keywords Genetic Algorith...

• Type: Conference
• Year: 2007
• Where: GECCO
• Authors: Justin Schonfeld
• Keywords: Algorithms, GECCO 2007, Random Sampling, Tournament Selection