Title :
The Power of Linear Programming for Valued CSPs
Author :
Thapper, Johan ; Zivny, S.
Author_Institution :
Lab. d´´Inf. (LIX), Ecole Polytech., Palaiseau, France
Abstract :
A class of valued constraint satisfaction problems (VCSPs) is characterised by a valued constraint language, a fixed set of cost functions on a finite domain. An instance of the problem is specified by a sum of cost functions from the language with the goal to minimise the sum. This framework includes and generalises well-studied constraint satisfaction problems (CSPs) and maximum constraint satisfaction problems (Max-CSPs). Our main result is a precise algebraic characterisation of valued constraint languages whose instances can be solved exactly by the basic linear programming relaxation. Using this result, we obtain tractability of several novel and previously widely-open classes of VCSPs, including problems over valued constraint languages that are: (1) sub modular on arbitrary lattices, (2) bisubmodular (also known as k-sub modular) on arbitrary finite domains, (3) weakly (and hence strongly) tree-sub modular on arbitrary trees.
Keywords :
algebra; constraint handling; linear programming; trees (mathematics); Max-CSP; VCSP; arbitrary finite domains; arbitrary lattices; arbitrary trees; linear programming relaxation; maximum constraint satisfaction problems; precise algebraic characterisation; valued constraint language; valued constraint satisfaction problems; Cloning; Complexity theory; Cost function; IP networks; Lattices; Linear programming; Robustness; bisubmodularity; fractional homomorphisms; fractional polymorphisms; linear programming; submodularity; valued constraint satisfaction;
Conference_Titel :
Foundations of Computer Science (FOCS), 2012 IEEE 53rd Annual Symposium on
Conference_Location :
New Brunswick, NJ
Print_ISBN :
978-1-4673-4383-1
DOI :
10.1109/FOCS.2012.25