Differential evolution algorithm using stochastic mutation

Differential Evolution (DE) Algorithm is a population-based meta-heuristic for solving various complex optimization problems. Mutation scale factor and Crossover constant are the two important control parameters of the algorithm which are used to direct the search process to the global optima. Fine tuning of both the parameters controls the performance of the algorithm. Literature suggests that due to large step sizes, DE is unable to exploit the promising search space. So, to improve exploitation capability of the algorithm, by taking inspiration from levy flight random walk, a new mutation scale factor, namely stochastic mutation factor is proposed and incorporated with DE. The proposed strategy is named as Differential Evolution Algorithm using Stochastic Mutation (DESMU). Further, to increase the exploration capability of the algorithm, a limit is associated with every solution to count the number of not updating iterations. If this count crosses the pre-defined limit then the solution is randomly initialized. The proposed algorithm is tested over 15 benchmark test functions and compared with basic (DE), and two of its variants namely Scale Factor Local Search Differential Evolution (SFLSDE) and Self-adaptive Differential Evolution (SADE). The results reveal that DESMU is a competitive variant of DE.