An Improved Estimation of Distribution Algorithms Based on the Minimal Free Energy

Abstract

To systematically harmonize the conflict between selective pressure and population diversity in estimation of distribution algorithms, an improved estimation of distribution algorithms based on the minimal free energy (IEDA) is proposed in this paper. IEDA conforms to the principle of minimal free energy in simulating the competitive mechanism between energy and entropy in annealing process, in which population diversity is measured by similarity entropy and the minimum free energy is simulated with an efficient and effective competition by free energy component. Through solving some typical numerical optimization problems, satisfactory results were achieved, which showed that IEDA was a preferable algorithm to avoid the premature convergence effectively and reduce the cost in search to some extent.