Tutorial: Niching Methods for Multimodal Optimization

Finding sub-optimal solutions in real-world problems can be challenging, and it can be beneficial to employ optimization strategies that seek multiple good solutions to increase the probability of finding a better solution. Niching techniques are often used to achieve this diversity of solutions and reveal richer information about the problem-domain. This tutorial provides an overview of niching methods for multimodal optimization, including traditional techniques and those specific to Particle Swarm Optimization and Differential Evolution. The tutorial offers a variety of real-world applications, from structural engineering to drug molecule design, and discusses benchmark suites and competitions used to evaluate and compare niching strategies. The tutorial highlights two recent successful strategies, utilizing Covariance Matrix self-adaptation evolution and hill-valley clustering, in benchmark problems for multimodal optimization. Additionally, niching techniques applied in clustering, feature selection, and dynamic optimization are also presented.