Stochastic fractal search with chaos

Abstract

In this study, the convergence speed and fitness function accuracy have been compared with the original algorithm by developing on the Stochastic Fractal Search (SFS) algorithm. Seven classical mathematical benchmark functions used in testing the optimization algorithms in the literature were used in comparison process. In the original SFS algorithm, the Gaussian walk function is used to find new solution points in diffusion process. The step length in this walk decreases as the iteration progresses and a function depending on generation value is used to provide for a more local search. The improvement in this work is the process of adding chaotic map values to this function. According to simulation results, it is observed that seven chaotic map improves the original algorithm from ten chaotic maps applied to SFS algorithm.