An Empirical Comparison of Differential Evolution Variants on Different Classes of Unconstrained Global Optimization Problems

摘要

This paper presents an empirical analysis of the performance of Differential Evolution (DE) variants on different classes of unconstrained global optimization benchmark problems. This analysis has been undertaken to identify competitive DE variants which perform reasonably well on a range of problems with different features. Towards this, fourteen DE variants were implemented and tested on 14 high dimensional benchmark functions grouped by their modality and decomposability viz., unimodal separable, unimodal nonseparable, multimodal separable and multimodal nonseparable. This extensive performance analysis provides some insight about the competitiveness of DE variants in solving test problems with representative landscape features such as modality and decomposability.