A new GA-Local Search Hybrid for Continuous Optimization Based on Multi-Level Single Linkage Clustering

Hybrid algorithms formed by the combination of Genetic Algorithms with Local Search methods provide increased performances when compared to real coded GA or Local Search alone. However, the cost of Local Search can be rather high. In this paper we present a new hybrid algorithm which reduces the total cost of local search by avoiding the start of the method in basins of attraction where a local optimum has already been discovered. Additionally, the clustering information can be used to help the maintenance of diversity within the population.