Inverse Mutations: Making the Evolutionary-Gradient-Search Procedure Noise Robust

R. Salomon (Germany)

Keywords

Evolutionary Algorithm, Convergence, Noisy Fitness Eval uation, EvolutionaryGradientSearch Procedure.

Abstract

Recent advances in the theory of evolutionary algo rithms have indicated that a hybrid method known as the evolutionary-gradient-search procedure yields supe rior performance in comparison to contemporary evolu tion strategies. But the theoretical analysis also indicates a noticeable performance loss in the presence of noise (i.e., noisy fitness evaluations). This paper aims at understand ing the reasons for this observable performance loss. It also proposes some modifications, called inverse mutations, to make the process of estimating the gradient direction more noise robust.

Important Links:



Go Back