Optimization of Multi-Extreme Multidimensional Functions: Population-Based Nature-Inspired Algorithm

Abstract

The article presents a population-based algorithm for solving multidimensional optimization problems using a hierarchical multi-population approach. Specific operators are used to supporting a diverse population of solutions, as well as expanding the search space by less promising solutions. The authors evaluate the effectiveness of the proposed algorithm on a set of multivariate functions of Schwefel, Rosenbrock, Rastrigin, and Grivank. The performance of the developed algorithm is compared with the performance of competing algorithms. The researchers register here significant statistical differences. In its turn, it proves in favor of a scalable evolutionary algorithm. Such a tendency is observed for all the considered functions with an increasing dimension of the problem.