Solving partial constraint satisfaction problems with tree decomposition
Arie M. C. A. Koster, Stan P.M. van Hoesel, Antoon W. J. Kolen · Networks · 2002
Abstract In this paper, we describe a computational study to solve hard partial constraint satisfaction problems (PCSPs) to optimality. The PCSP is a general class of problems that contains a diversity of problems, such as generalized subgraph problems, MAX‐SAT, Boolean quadratic programs, and assignment problems like coloring and frequency planning. We present a dynamic programming algorithm that solves PCSPs based on the structure (tree decomposition) of the underlying constraint graph. With the use of dominance and bounding techniques, we are able to solve small and medium‐size instances of the problem to optimality and to obtain good lower bounds for large‐size instances within reasonable time and memory limits. © 2002 Wiley Periodicals, Inc.