Abstract
Differential evolution (DE) is a versatile and fast evolutionary algorithm (EA) for real world global optimization problems. It has been widely applied to diverse areas as a remarkable search technique. However, because of large step sizes in mutation, DE is known to be incapable of exploiting the existing population than exploring the search region. DE inherent drawback is avoiding the true optima. To heal its poor exploitation, hybridization with local search techniques will be a healthy choice, because local search strategies are proven to be robust in exploitation. In this paper, we hybridize DE variant, JADE with two gradient based local search (LS) techniques, Steepest Decent Method (SDM) and Broyden-Fletcher-Goldfarb-Shanno (BFGS) method. The new algorithm is known as hybidJADE. The SDM is applied iteratively to locate promising solutions in the evolution, afterwards BFGS is applied to fine tune the elite solutions. If BFGS found solutions are in the neighborhood of current best solution, a restart is incorporated, where JADE and BFGS explore the population. The performance of hybridJADE is tested on 15 benchmarks from literature. The obtained experimental results are compared with the results of two state-of-the-art algorithms, JADE, and jDE. The results show significance performance of hybridJADE in terms of success rate on various benchmarks. The sensitivity analysis of hybridJADE to its various parameters is also presented.