Evolution strategies for objective functions with locally correlated variables

Abstract

This paper proposes an improvement of Evolutionary Strategies for objective functions with locally correlated variables. It focusses on detecting local dependencies among variables of the objective function on the basis of the current population and transforming the original objective function into a new one of a smaller number of variables. Such a transformation is updated in successive iterations of the evolutionary algorithm to reflect local dependencies over successive neighborhoods of optimal solutions. Experiments performed on some popular benchmark functions confirm that the improved algorithm outperforms the original one. © 2010 Springer-Verlag Berlin Heidelberg.