抄録
Graph partitioning is an important problem that has extensive applications in many areas, including VLSI design, scientific computing, data mining, geographical information systems and job scheduling. The graph partitioning problem (GPP) is NP-complete. There are several heuristic algorithms developed finding a reasonably good resolution. The most famous partitioning methods are simulated annealing (SA) and mean field algorithm (MFA) known to produce good partition for a wide class of problems, and they are used quite extensively. However these methods are very expensive in time and very sensitive in parameters tuning methods.
In this paper, a new parameter-free algorithm for GPP has been proposed. The algorithm has been constructed using the S-model learning automata with multi-teacher random environments. As shown in our experiments, the proposed algorithm has some advantages superior to SA, MFA and ParMeTiS.
