Volume 5, Number 2, May 2009, pp. 227-236

Satoko Moriguchi and Nobuyuki Tsuchimura
Key words:
discrete optimization, discrete convex function, submodular function, algorithm
Mathematices Subject Classification: 52A41, 90C27
ONLINE SUBSCRIPTION (Institutional Subscription Only)
Copyright© 2008 Yokohama Publishers
Back

Abstract:
We consider the problem of minimizing a nonlinear discrete function with -convexity proposed in the theory of discrete convex analysis. For this problem, a steepest descent algorithm and a steepest descent scaling algorithm are known. In this paper, we propose a continuous relaxation approach which first minimizes the continuous variable version in order to find a good initial solution of the steepest descent algorithm. For discrete -convex functions, we give a proximity theorem showing that a discrete global minimizer exists in a neighborhood of a continuous global minimizer. This proximity theorem affords a theoretical guarantee for the efficiency of the proposed algorithm.
Discrete L-convex function minimization based on continuous relaxation