Length-lex ordering for set CSPs

  • Authors:
  • Carmen Gervet;Pascal Van Hentenryck

  • Affiliations:
  • Brown University, Providence, RI;Brown University, Providence, RI

  • Venue:
  • AAAI'06 Proceedings of the 21st national conference on Artificial intelligence - Volume 1
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

Combinatorial design problems arise in many application areas and are naturally modelled in terms of set variables and constraints. Traditionally, the domain of a set variable is specified by two sets (R,E) and denotes all sets containing R and disjoint from E. This representation has inherent difficulties in handling cardinality and lexicographic constraints so important in combinatorial design. This paper takes a dual view of set variables. It proposes a representation that encodes directly cardinality and lexicographic information, by totally ordering a set domain with a length-lex ordering. The solver can then enforce bound-consistency on all unary constraints in time Õ(k) where k is the set cardinality. In analogy with finite-domain solvers, non-unary constraints can be viewed as inference rules generating new unary constraints. The resulting set solver achieves a pruning (at least) comparable to the hybrid domain of Sadler and Gervet at a fraction of the computational cost.