Valuated matroid-based algorithm for submodular welfare problem

Takanori Maehara, Kazuo Murota

Abstract: An algorithm for the submodular welfare problem is proposed based on the theory of discrete convex analysis. The proposed algorithm is a heuristic method built upon the valuated matroid partition algorithms, and gives the exact optimal solution for a reasonable subclass of submodular welfare problems. The algorithm has a guaranteed approximation ratio for a special case. Computational results show fairly good performance of the proposed algorithm.