Levy Flight Based Local Search in Differential Evolution

摘要

In order to solve non convex and complex optimization problems, Nature Inspired algorithms are being preferred in present scenario. Differential Evolution (DE) is relatively popular and simple population based probabilistic algorithm under the said category to find optimum value. The scale factor (F) and crossover probability (CR) are the two parameters which controls the performance of DE in its mutation and crossover processes by maintaining the balance between exploration and exploitation in search space. Literature suggests that due to large step sizes, DE is less capable of exploiting the existing solutions than the exploration of search space. Therefore unlike the deterministic methods, DE has inherent drawback of skipping the true optima. This paper incorporates the Levy Flight inspired local search strategy with DE named as Levy Flight DE (LFDE) which exploits the search space identified by best solution. To see the performance of LFDE, experiments are carried out on 15 benchmark problems of different complexities and results show that LFDE is a competitive DE variant and perform better than the basic DE and its recent variants namely Fitness based DE (FBDE) and Scale Factor Local Search DE (SFLSDE) in most of the test functions.