Résumé
Virtual Arc Consistency (VAC) is a recent local consistency for processing Cost Function Networks (or Weighted Con- straint Networks) that exploits a simple but powerful con- nection with classical Constraint Networks. It has allowed to close hard frequency assignment benchmarks and is ca- pable of directly solving networks of submodular functions. The algorithm enforcing VAC is an iterative algorithm that solves a sequence of classical Constraint Networks. In this work, we show that Dynamic Arc Consistency algorithms can be suitably injected in the virtual arc consistency itera- tive algorithm, providing noticeable speedups.