On division versus saturation in pseudo-boolean solving

Forskningsoutput: Kapitel i bok/rapport/Conference proceedingKonferenspaper i proceeding

Abstract

The conflict-driven clause learning (CDCL) paradigm has revolutionized SAT solving over the last two decades. Extending this approach to pseudo-Boolean (PB) solvers doing 0-1 linear programming holds the promise of further exponential improvements in theory, but intriguingly such gains have not materialized in practice. Also intriguingly, most PB extensions of CDCL use not the division rule in cutting planes as defined in [Cook et al.,'87] but instead the so-called saturation rule. To the best of our knowledge, there has been no study comparing the strengths of division and saturation in the context of conflict-driven PB learning, when all linear combinations of inequalities are required to cancel variables. We show that PB solvers with division instead of saturation can be exponentially stronger. In the other direction, we prove that simulating a single saturation step can require an exponential number of divisions. We also perform some experiments to see whether these phenomena can be observed in actual solvers. Our conclusion is that a careful combination of division and saturation seems to be crucial to harness more of the power of cutting planes.

Detaljer

Författare
Externa organisationer
  • KTH Royal Institute of Technology
  • University of Copenhagen
  • Technion - Israel Institute of Technology
Forskningsområden

Ämnesklassifikation (UKÄ) – OBLIGATORISK

  • Datavetenskap (datalogi)
Originalspråkengelska
Titel på värdpublikationProceedings of the 28th International Joint Conference on Artificial Intelligence, IJCAI 2019
RedaktörerSarit Kraus
FörlagInternational Joint Conferences on Artificial Intelligence
Sidor1711-1718
Antal sidor8
ISBN (elektroniskt)9780999241141
StatusPublished - 2019
PublikationskategoriForskning
Peer review utfördJa
Externt publiceradJa
Evenemang28th International Joint Conference on Artificial Intelligence, IJCAI 2019 - Macao, Kina
Varaktighet: 2019 aug 102019 aug 16

Publikationsserier

NamnIJCAI International Joint Conference on Artificial Intelligence
Volym2019-August
ISSN (tryckt)1045-0823

Konferens

Konferens28th International Joint Conference on Artificial Intelligence, IJCAI 2019
LandKina
OrtMacao
Period2019/08/102019/08/16