Journal "Software Engineering"
a journal on theoretical and applied science and technology
ISSN 2220-3397
Journal "Software Engineering"
a journal on theoretical and applied science and technology
ISSN 2220-3397
Issue N10 2019 year
Evolutionary algorithms have particular advantages over traditional deterministic optimization methods. For example, class of algorithms better explores the solution space. Moreover, big data and multidimensional problems add additional complexity when we want to explore the solution search space. The paper presents an evolutionary algorithm capable of solving multidimensional optimization problems of high dimension. The algorithm uses a hierarchical multi-population approach, as well as special operators to support the diversity of the decision population and to expand the area of finding solutions with less promising solutions. The efficiency of the proposed algorithm is evaluated on a set of multidimensional optimization functions-benchmarks of Griewank, Rastrigin, Rosenbrock, and Schwefel. Moreover, the indicators of the developed algorithm are compared with the indicators of competing algorithms. We can note statistically significant differences. In its turn, this fact confirms the advantages of a scalable evolutionary algorithm for all considered benchmark functions especially with an increasing dimension of the problem. The authors believe that this is due to the ability of the algorithm to maintain the diversity of the population and to find a balance between the rate of convergence of the algorithm and the diversification of the search. To sum up, the algorithm is a quite promising way for solving multidimensional optimization problems with complex multimodal solution spaces and big data.