An efficient simulated annealing algorithm for graph bisectioning

A new simulated annealing algorithm for solving the graph bisectioning problem is proposed. The authors run their simulated annealing algorithm, the Kernighan-Lin algorithm, and the Saab-Rao algorithms on the same set of random graphs with 50 to 500 nodes and compare their performances. Experiments show that their simulated annealing algorithm provides lower bisection cost than the Kernighan-Lin algorithm and the Saab-Rao algorithms for all of the graphs and their algorithm takes less running time than the other algorithms mentioned for all of the graphs with more than 100 nodes. For the simulated annealing approach, they conclude that sequential neighborhood search outperforms random neighborhood search in solving the graph bisectioning problem.<
>