Coordinatewise domain scaling algorithm for M-convex function minimization

We present a polynomial time scaling algorithm for the minimization of an M-convex function. M-convex functions are nonlinear discrete functions with (poly)matroid structures, which are being recognized as playing a fundamental role in tractable cases of discrete optimization. The algorithm is applicable also to a variant of quasi M-convex functions.Mathematics Subject Classification (2000):90C27, 68W40, 05B35This work is supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan.