A hybrid metaheuristic based on differential evolution and local search with quadratic interpolation

Abstract

The use of Local Search technique in combination with other methods is often an effective way for increasing the e ciency of a global optimization algorithm. In this paper we present an hybrid version that integrates Di erential Evolution with Local Search, applying the Quadratic Interpolation formula for determining the neighborhood in which to explore towards better solutions. We present DE+LS(1) in which the closer neighborhood to the best population individual is explored, and DE+LS(2) in which the neigh- borhood of the two best population individuals is examined. The results showed that with DE+LS(2) improvements are not signi cant, but using DE+LS(1) an improvement is achieved, especially for large dimensions, in terms of solutions quality and speed of convergence.