Introducing Self-Adaptive Parameters to Self-organizing Migrating Algorithm

In this paper, a new modification for a modern and popular optimization Self Organizing Migrating Algorithm (SOMA) is presented. SOMA resembles swarm-based algorithms together with mutation process given by perturbation and self-adaptation of individual's migration over the hyperspace of a given optimized solution. However, the quality of the solution found by SOMA strongly depends on user-defined parameters. This is not problematic only for new users, but sometimes for experts as well. The proposed modification allows individual (solution) to change its parameters based on its actual performance and adapts to specific optimization problems. The recent CEC'17 benchmark suite is used for analyzing an original SOMA and performance testing of a proposed SOMA modification. The results are compared and tested for statistical significance.