The role of explicit niching and communication messages in distributed evolutionary multi-objective optimization

Abstract

This chapter is dedicated to an investigation on the role of explicit niching and communication messages in distributed evolutionary multi-objective optimization. Localization is employed to implement explicit niching. Several options are selected for communication messages including non-dominated solutions and statistics such as the centroid of the non-dominated set, the direction of improvement, or weighted direction of improvement. As a result, a distributed system using the framework of local models is developed to support distributed computing in evolutionary multi-objective optimization. This system provides a flexibility in applying different architectures such as master/slave, island as well as the hybridization of the two. An in-depth analysis is carried out on a simulation study using the system. © 2010 Springer-Verlag Berlin Heidelberg.