DMCA
GENERALIZED CONSTRAINT SATISFACTION PROBLEMS (2006)
| Citations: | 1 - 1 self |
BibTeX
@MISC{Scott06generalizedconstraint,
author = {Alexander D. Scott and Gregory and B. Sorkin},
title = {GENERALIZED CONSTRAINT SATISFACTION PROBLEMS},
year = {2006}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract. A number of recent authors have given exponential-time algorithms for optimization problems such as Max Cut and Max Independent Set, or for the more general class of Constraint Satisfaction Problems (CSPs). In this paper, we introduce the class of Generalized Constraint Satisfaction Problems (GCSPs), where the score functions are polynomial-valued rather than realvalued functions. We show that certain reductions used for solving CSPs can be extended to identities for the “partition function ” of a GCSP, leading to relatively efficient exponential-time (polynomial-space) algorithms for solving a GCSP. This also enables us (at the cost of only a polynomial factor in time) to modify existing algorithms for optimizing CSPs into algorithms that count solutions or sample uniformly at random. Using an extra variable allows us to solve Max Bisection or calculate the partition function of the Ising Model, problems that were previously inaccessible with this approach. 1.
Keyphrases
generalized constraint satisfaction problem partition function max cut extra variable optimization problem recent author exponential-time algorithm polynomial factor general class max bisection certain reduction score function max independent set ising model constraint satisfaction problem
