The complexity of complex weighted Boolean #CSP

  • Authors:
  • Jin-Yi Cai;Pinyan Lu;Mingji Xia

  • Affiliations:
  • Computer Sciences Department, University of Wisconsin, Madison, WI 53706, USA;Microsoft Research Asia, Beijing, 100080, PR China;Max-Planck-Institut für Informatik, 66123 Saarbrücken, Germany

  • Venue:
  • Journal of Computer and System Sciences
  • Year:
  • 2014

Quantified Score

Hi-index 0.00

Visualization

Abstract

We prove a complexity dichotomy theorem for the most general form of Boolean #CSP where every constraint function takes values in the field of complex numbers C. We first give a non-trivial tractable class of Boolean #CSP which was inspired by holographic reductions. The tractability crucially depends on algebraic cancelations which are absent for non-negative numbers. We then completely characterize all the tractable Boolean #CSP with complex-valued constraints and show that we have found all the tractable ones, and every remaining problem is #P-hard. We also improve our result by proving the same dichotomy theorem holds for Boolean #CSP with maximum degree 3 (every variable appears at most three times). The concept of Congruity and Semi-congruity provides a key insight and plays a decisive role in both the tractability and hardness proofs. We also introduce local holographic reductions as a technique in hardness proofs.