A novel compact genetic algorithm using offspring survival evolutionary strategy

Abstract

This article describes a compact genetic algorithm (cGA) with an offspring survival evolutionary strategy. The cGA requires less memory than the population-based GA since the whole population is not necessary. The cGA can easily be implemented because it has no complex genetic operator. However, the cGA requires a large amount of fitness evaluation to provide acceptable solutions in problems involving higher-order building blocks (BBs). In order to reduce the number of fitness evaluations, a higher selection pressure is applied to the cGA. Generally, elitism is used to increase the selection pressure. However, elitism may lead to premature convergence as the order of BBs becomes higher. In this article, we propose a balanced cGA using an offspring survival evolutionary strategy. The usefulness of the proposed cGA is verified by comparing it with the original cGA and the elitism-based cGAs using wellknown benchmark functions.