Constrained Optimization Using Triple Spaces Cultured Genetic Algorithm

Abstract

Cultural algorithms provide a useful framework to make evolutionary algorithms more efficient. However, there is still much for revision especially when they are applied in the constrained optimizations, where a mass of memory and computation cost is currently unavoidable. We propose a novel triple spaces cultural algorithm in which a new framework called anti-culture population consisting of individuals disobeying the guidance of culture is added to the traditional dual inheritance cultural algorithm. The effect that the individuals in the anti-culture population disobey culture's guidance is ensured by some mutation operations which make the individuals away from the Culture guided individual in a radiating way. The anti-culture population makes the evolution of both culture and the population faster and at the same time take a lower risk of the local optimization problem. Moreover, with the triple spaces structure and some novel rules to control the convergence process of the algorithm through awarding the most successful individuals and punishing the unsuccessful population, it is possible to deal with a constrained optimization problem with computation burden almost the same as that in solving unconstrained optimization problems. genetic algorithm is utilized as the basis of the population space due to its advantages in representing the structure of the space and convenience in computation. Comparisons with four reported algorithms show that our proposed approach has significant advantages while the cost of computation and storage is much lower.