Constrained optimization problem solving using estimation of distribution algorithms

Abstract

Two variants of estimation of distribution algorithm (EDA) are tested solving several continuous optimization problems with constraints. Numerical experiments are conducted and comparison is made between constraint handling using several types of penalty and repair operators in case of both elitist and nonelitist implementation of the EDA's. Graphical display and animations of representative runs of the best and worst performers proved useful in enhancing the understanding of how such algorithms work.