A Simple and Effective Termination Condition for Both Single- and Multi-Objective Evolutionary Algorithms

Abstract

In this paper, a simple and effective termination condition for both single- and multi-objective evolutionary algorithms has been proposed. The termination condition is based on simply observing objective values of solution candidates during generations. Effectiveness of the termination condition is self-evident with single-objective problems but unclear with multi-objective problems. Therefore, experiments with some well known bi- and tri-objective test problems have been performed. The proposed termination condition is implemented in Generalized Differential Evolution (GDE) that is a general purpose optimization algorithm for both single- and multi-objective optimization with or without constraints. Our preliminary results indicate that the proposed termination condition is a suitable termination condition also with multi-objective problems. With the termination condition and a control parameter adaptation technique previously introduced, GDE has become a fully automated optimization algorithm that can be used by any optimization practitioner.