EDA-PSO: A hybrid paradigm combining estimation of distribution algorithms and particle swarm optimization

Abstract

Estimation of Distribution Algorithms (EDAs) is an evolutionary computation optimization paradigm that relies the evolution of each generation on calculating a probabilistic graphical model able to reflect dependencies among variables out of the selected individuals of the population. This showed to be able to improve results with GAs for complex problems. This paper presents a new hybrid approach combining EDAs and particle swarm optimization, with the aim to take advantage of EDAs capability to learn from the dependencies between variables while profiting particle swarm's optimization ability to keep a sense of "direction" towards the most promising areas of the search space. Experimental results show the validity of this approach with widely known combinatorial optimization problems. © 2010 Springer-Verlag Berlin Heidelberg.