Effect of a Push Operator in Genetic Algorithms for Multimodal Optimization

Genetic Algorithms have been successfully used to solve multimodal optimization problems, mainly due to their population approach and implicit parallel processing among multiple subpopulations. In order to find and maintain multiple regions, GAs implement a niching principle motivated from nature. The selection procedure of a GA is modified by restricting a comparison among similar solutions to bring about an additional level of diversity in the population. In another recent study, a real-parameter push-operator based on non-uniform coding principle applied to binary-coded GAs was proposed. The push-operator has shown to exhibit better convergence properties on many optimization problems compared to standard GA implementations. In this paper, we extend the push-operator and its implementation with the niching principle to solve multi-modal problems. On a number of constrained and unconstrained multi-modal test problems, we demonstrate its superior convergence to multiple optimal solutions simultaneously. Results are interesting and motivate us to extend the push-operator to multi-objective and other complex optimization tasks.