Primal-dual genetic algorithms for royal road functions

Abstract

Based on Holland's simple genetic algorithm (SGA) three have been many variations developed. Inspired by the phenomenon of diploid genotype and deminance mechanisms broadly existing in nature, we have proposed a primal-dual genetic algorithm (PDGA), see (Yang 2002). Our preliminary experiments based on the Royal Road functions have shown that PDGA outperforms SGA for different performance measures. In this paper, we present some further experiment results, especially onthe dynamic performance of PDGA over SGA, and give out our explanations and analyses about ehy PDGA outperforms SGA based on these results. Through the primal-dual mapping between a pair of chromosomes, PDGA's performance of exploration in the search space, especially during the early generations, is improved and thus its total searching efficiency is improved.