COSATS, X-COSATS: Two multi-agent systems cooperating simulated annealing, tabu search and X-over operator for the K-Graph partitioning problem

Abstract

In this paper we propose two multi-agent systems gathering several metaheuristics for the K-Graph Partitioning Problem(K-GPP). In the first model COSATS, two metaheuristic agents, namely Tabu Search and Simulated Annealing run simultaneously to solve the K-GPP. These agents are mutually guided during their search process by means of a new mechanism of information exchange based on statistical analysis of search space. In the second model X-COSATS, a crossover agent is added to the model in order to make a cross-over between the local optima found by simulated annealing and tabu agents. COSATS and X-COSATS are tested on several large graph benchmarks. The experiments demonstrated that our models achieve partitions with significantly higher quality than those generated by simulated annealing and tabu search operating separately © Springer-Verlag Berlin Heidelberg 2005.