An Overview of Evolutionary Computing for Multimodal Function Optimisation

Abstract

Evolutionary Computing (EC) is becoming popular among researchers in multimodal function optimisation (MFO). One of the major issues in MFO is maintaining the ‘useful’ diversity in the population. The ‘useful’ diversity is utilised in search either to achieve the global optimum or to maintain multiple sub-optima in the final population. In case of multimodal functions these two goals can be dependent on each other. The paper discusses the principle issues involved in MFO. All major EC techniques suitable for MFO are categorised, and their strengths and weaknesses are discussed. Application of EC techniques to solve real life MFO problems are difficult. Real life problems can pose some additional challenge than test functions. Real life problems are difficult mainly because of lack of prior knowledge. The paper mentions the challenges posed by such real life problems, and discusses how Adaptive Restricted Tournament Selection (ARTS) can address some of the challenges. From the discussion, direction for future research in MFO is identified.