Enhancing a Multi-population Optimisation Approach with a Dynamic Transformation Scheme

Abstract

The adaptive multi-population optimisation (AMPO) algorithm is an intelligent meta-heuristic search method utilising multiple search groups to conduct a diversity of search strategies in evolutionary algorithms or swarm intelligence. With the careful design of different search operators, the AMPO algorithm has achieved outstanding performance in many optimisation problems including two sets of benchmark functions when compared to some latest approaches including the hybrid firefly and particle swarm optimisation for continuous optimisation. Yet there are still opportunities to enhance the adaptability of its search mechanism in various aspects. Therefore, a more adaptive AMPO (AMPO$^{+}$) algorithm is considered in this work in which the probability of the transformation between specific search groups can be more flexibly adjusted during the different stages of the search process. In this way, the AMPO$^{+}$ can better adapt its search efforts to specific search groups through revising its search strategies so as to effectively solve many challenging optimisation problems. To carefully examine the search effectiveness of the enhanced framework, the proposed AMPO$^{+}$ algorithm is evaluated against the original AMPO and other sophisticated meta-heuristic algorithms on a set of well-known benchmark functions of different dimensions in which impressive results are attained by the AMPO$^{+}$. More importantly, the proposed adaptive search framework sheds light on many possible directions for further investigation.