## Answer set programming with clause learning (2004)

Venue: | In: LPNMR-7. LNCS, (2004) 302–313 F. Ricca |

Citations: | 23 - 0 self |

### BibTeX

@INPROCEEDINGS{Ward04answerset,

author = {Jeffrey Ward and John S. Schlipf},

title = {Answer set programming with clause learning},

booktitle = {In: LPNMR-7. LNCS, (2004) 302–313 F. Ricca},

year = {2004},

pages = {302--313},

publisher = {Springer}

}

### Years of Citing Articles

### OpenURL

### Abstract

Abstract. A conflict clause represents a backtracking solver’s analysis of why a conflict occurred. This analysis can be used to further prune the search space and to direct the search heuristic. The use of such clauses has been very important in improving the efficiency of satisfiability (SAT) solvers over the past few years, especially on structured problems coming from applications. We describe how we have adapted conflict clause techniques for use in the answer set solver Smodels. We experimentally compare the resulting program to the original Smodels program. We also compare to ASSAT and Cmodels, which take a different approach to adding clauses to constrain an answer set search. 1

### Citations

1192 | Chaff: engineering an efficient SAT solver
- Moskewicz, Madigan, et al.
(Show Context)
Citation Context ...plications seems to be that some variables, and some combinations of variables, are far more important than others. More recent SAT solvers such as GRASP [MS99], SATO [Zheng97], rel sat [BS97], Chaff =-=[MMZZM01]-=-, BerkMin [GN02], and SIMO [GMT03], through creating and processing conflict clauses, often “learn” important information for the search. The Cmodels–1 solver [Cmod] addresses this problem by piggy-ba... |

977 |
Negation as Failure
- Clark
- 1978
(Show Context)
Citation Context ...backing an answer set solver onto a SAT solver. It handles a class of logic programs called tight [Fages94, BEL00], in which the stable models are just the models of Clark’s completion of the program =-=[Clark78]-=- — and the completion is a classical logic problem. So Cmodels– 1, after some preprocessing, passes a completion to a SAT solver, such as Chaff. Our concern in this paper is with solvers which are not... |

769 |
A machine program for theorem proving
- Davis, Logemann, et al.
- 1962
(Show Context)
Citation Context ... set calculation directly into the search engine, thus allowing tighter pruning of the search tree. 2 Background Propositional CNF-SAT solvers. A basic Davis-Putnam-Loveland-Logeman (DPLL) SAT solver =-=[DLL62]-=-, given a set C of clauses, performs a backtracking search to find a model for C. We sketch it below as a recursive function, passed a set of literals representing a partial truth assignment. Initiall... |

399 | GRASP: A Search Algorithm for Propositional Satisfiability
- Marques-Silva, Sakallah
- 1999
(Show Context)
Citation Context ...cient data structures. Key to many of these applications seems to be that some variables, and some combinations of variables, are far more important than others. More recent SAT solvers such as GRASP =-=[MS99]-=-, SATO [Zheng97], rel sat [BS97], Chaff [MMZZM01], BerkMin [GN02], and SIMO [GMT03], through creating and processing conflict clauses, often “learn” important information for the search. The Cmodels–1... |

335 | Logic programs with stable model semantics as a constraint programming paradigm
- Niemela
- 1999
(Show Context)
Citation Context ...not tight. We used directed graphs in these experiments. Here we considered three reductions to answer set programming. The first, a standard reduction frequently used in benchmarking, was taken from =-=[Nie99]-=-. 8 The second reduction was a “tight on its models” reduction used with ASSAT in [LZ03]. The third reduction is the modification below of the first: dlx2 cc bug01.cnf, ..., dlx2 cc bug08.cnf The eigh... |

320 | Extending and implementing the stable model semantics - Simons, Niemelä, et al. |

306 | Efficient conflict driven learning in a Boolean satisfiability solver
- Zhang, Madigan, et al.
(Show Context)
Citation Context ...t with the unfounded set check and thus sacrifices this modularity. 3 Conflict Clause Generation Crucial to many modern SAT solvers is creating, storing, and looking up conflict clauses. Zhang et al. =-=[ZMMM01]-=- studies several different strategies for generating conflict clauses, implementing each of them in the zChaff variant of Chaff. We describe here their most effective strategy (the 1UIP strategy) and ... |

218 | Using csp look-back techniques to solve real-world sat instances
- Bayardo, Schrag
- 1997
(Show Context)
Citation Context ...ny of these applications seems to be that some variables, and some combinations of variables, are far more important than others. More recent SAT solvers such as GRASP [MS99], SATO [Zheng97], rel sat =-=[BS97]-=-, Chaff [MMZZM01], BerkMin [GN02], and SIMO [GMT03], through creating and processing conflict clauses, often “learn” important information for the search. The Cmodels–1 solver [Cmod] addresses this pr... |

211 | Assat: Computing answer set of a logic program by sat solvers - Lin, Zhao - 2004 |

207 | An efficient propositional prover
- ZHANG
(Show Context)
Citation Context ...ructures. Key to many of these applications seems to be that some variables, and some combinations of variables, are far more important than others. More recent SAT solvers such as GRASP [MS99], SATO =-=[Zheng97]-=-, rel sat [BS97], Chaff [MMZZM01], BerkMin [GN02], and SIMO [GMT03], through creating and processing conflict clauses, often “learn” important information for the search. The Cmodels–1 solver [Cmod] a... |

147 | Consistency of Clark’s completion and existence of stable models - Fages - 1993 |

96 | Superscalar processor verification using efficient reductions of the logic of equality with uninterpreted functions
- Velev, Bryant
- 1999
(Show Context)
Citation Context ...trated on three problem domains: Boolean satisfiability. Our tests in this domain include some randomly generated 3SAT problems and 16 of the “DLX” circuit verification benchmarks from Miroslav Velev =-=[VB99]-=-. 7 In each case, the problem is provided as a CNF-SAT problem, which we have converted to an answer set program. We do not expect to outperform ASSAT or 6 We also tested a version of Smodelscc which ... |

79 | Cmodels-2: Sat-based answer set solver enhanced to non-tight programs
- Lierler, Maratea
- 2004
(Show Context)
Citation Context ...ling a SAT solver, evaluating whether the model produced is stable, and, if not, adding further specifications (“loop formulas”) to the problem to avoid this failure of stability. Recently, Cmodels–2 =-=[BM03]-=- adapts the technique of ASSAT, but extends its application to extended and disjunctive rules, and makes available alternative loop formula definitions. Here we present an answer set programming tool ... |

64 | Declarative problem-solving using the DLV system
- Eiter, Faber, et al.
- 2000
(Show Context)
Citation Context ...ent approach to adding clauses to constrain an answer set search. 1 Introduction Recent years have seen the development of several stable model/answer set solvers. Smodels [Sim00,NSS00,Nie99] and DLV =-=[EFLP00]-=- are commonly used. (DLV implements more, namely disjunctive logic programming. However, it also serves as an effective stable model search engine.) These have demonstrated the feasibility of answer s... |

44 | Smodels: A system for answer set programming - Niemelä, Simons, et al. - 2000 |

31 | Fages’ theorem and answer set programming - Babovich, Erdem, et al. - 2000 |

30 |
BerkMin: A Fast and Robust
- Goldberg, Novikov
- 2002
(Show Context)
Citation Context ... be that some variables, and some combinations of variables, are far more important than others. More recent SAT solvers such as GRASP [MS99], SATO [Zheng97], rel sat [BS97], Chaff [MMZZM01], BerkMin =-=[GN02]-=-, and SIMO [GMT03], through creating and processing conflict clauses, often “learn” important information for the search. The Cmodels–1 solver [Cmod] addresses this problem by piggy-backing an answer ... |

4 | Look-ahead vs. look-back techniques in a modern SAT solver
- Giunchiglia, Maratea, et al.
- 2003
(Show Context)
Citation Context ...iables, and some combinations of variables, are far more important than others. More recent SAT solvers such as GRASP [MS99], SATO [Zheng97], rel sat [BS97], Chaff [MMZZM01], BerkMin [GN02], and SIMO =-=[GMT03]-=-, through creating and processing conflict clauses, often “learn” important information for the search. The Cmodels–1 solver [Cmod] addresses this problem by piggy-backing an answer set solver onto a ... |

1 |
On Tight Logic Programs andYetAnother Translation from Normal Logic Programs to Propositional Logic
- Lin, Jicheng
- 2003
(Show Context)
Citation Context ...ons to answer set programming. The first, a standard reduction frequently used in benchmarking, was taken from [Nie99]. 8 The second reduction was a “tight on its models” reduction used with ASSAT in =-=[LZ03]-=-. The third reduction is the modification below of the first: dlx2 cc bug01.cnf, ..., dlx2 cc bug08.cnf The eight unsatisfiable instances were dlx1 c.cnf, dlx2 aa.cnf, dlx2 ca.cnf, dlx2 cc.cnf, dlx2 c... |

1 |
code is included in the Cmodels-2 distribution. See also http://www.mrg.dist.unige.it/ sim/simo
- Simo
(Show Context)
Citation Context ...in. The SAT solver used by ASSAT was Chaff2 (version spelt3) [Chaff2], which is ASSAT’s default solver. Cmodels used the zChaff SAT solver (version 2003.7.1) for tight problems and SIMO (version 3.0) =-=[SIMO]-=- for Hamiltonian cycle problems, as dictated by the command line settings. Note thatASSAT and Cmodels benefit from conflict clauses in our tests, because conflict clauses are heavily incorporated into... |