The baldwin effect on a memetic differential evolution for constrained numerical optimization problems
Saúl Domínguez-Isidro, Efrén Mezura‐Montes · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2017
This paper analyzes the Baldwin effect on a memetic algorithm that solves constrained numerical optimization problems (CNOPS). For this study the canonical Differential Evolution (DE) enhanced with the Hooke-Jeeves method (HJ) as local search operator is proposed (MDEHJ), which implements a probabilistic scheme to activate HJ by means a sinusoidal function that considers the population diversity. Three MDEHJ instances are applied to study the Baldwin effect in different exploitation areas (best, worst and random selected, respectively). Final results are compared against those obtained by MDEHJ with Lamarckian learning. All instances are tested on thirty-six well-known benchmark problems. The results suggest that the proposed approach is suitable to solve CNOPS and those results also show that Baldwin effect does not affect the performance of a memetic DE in constrained search spaces.