Properties of the island-based and single population differential evolution algorithms applied to discrete-continuous scheduling

Abstract

In this paper we have studied by experiment the properties of two models of the DE search: a model based on a single population, and a model based on multiple populations, known as the island model (IBDEA). We consider two versions of the island model: with migration of individuals between islands and without migration. We investigated how the effectiveness of models depends on such parameters as the size of a single population, and in the case of the island model, also the number of islands and migration rate between them. The general conclusion is that both models can be equally effective when used with proper parameter settings, which have been determined by the experiment. In addition, conditions for higher effectiveness of the IBDEA were discussed. The discrete-continuous scheduling with continuous resource discretisation was used as the test problem.