COSATS: A new cooperation model between simulated annealing and Tabu search for the K-graph partitioning problem

Abstract

In our model, two heuristic agents, namely Tabu Search and Simulated Annealing run simultaneously to solve the K-Graph Partitioning Problem (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. COSATS is tested on several large graph benchmarks. The experiments demonstrated that our model achieves partitions with significantly higher quality than those generated by simulated annealing and tabu search operating separately.