A GREEDY ALGORITHM FOR MINIMIZING A SEPARABLE CONVEX FUNCTION OVER AN INTEGRAL BISUBMODULAR POLYHEDRON

We present a new greedy algorithm for minimizing a separable convex function over an integral bisubmodular polyhedron. The algorithm starts with a.n arbitrary feasible solution and a current feasible solution incrementally moves toward an optimal one in a greedy way. \Ve also show that there exists at least one optimal solution in the coordinate-wise steepest descent direction from a feasible solution if it is not an optimal one.