Learning hybridization strategies in evolutionary algorithms
Intelligent Data Analysis
The differential ant-stigmergy algorithm
Information Sciences: an International Journal
Hi-index | 0.00 |
Evolutionary Algorithms (EAs) are powerful metaheuristics that can be applied to almost any optimization problem. However, different Evolutionary Algorithms own different search capabilities that make them more suitable for one or another optimization problem. Furthermore, the combination of several EAs can boost the performance of individual approaches. In this paper we try to exploit the benefits of the combination of several evolutionary approaches by means of a Hybrid Evolutionary Algorithm, where the participation of each individual algorithm on the overall process is dynamically adjusted through its execution. Two different strategies to perform this adjustment are proposed: one with a constant global population size and another one with variable global population size. Experimental results demonstrate that Hybrid Evolutionary Algorithms outperform the individual ones and that the dynamic strategy with variable population size obtains better results on most of the proposed functions.