Machine Learning-Based Restart Policy for CDCL SAT Solvers

被引:19
|
作者
Liang, Jia Hui [1 ]
Oh, Chanseok [2 ]
Mathew, Minu [3 ]
Thomas, Ciza [3 ]
Li, Chunxiao [1 ]
Ganesh, Vijay [1 ]
机构
[1] Univ Waterloo, Waterloo, ON, Canada
[2] Google, New York, NY USA
[3] Coll Engn, Thiruvananthapuram, Kerala, India
关键词
SATISFIABILITY;
D O I
10.1007/978-3-319-94144-8_6
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
Restarts are a critically important heuristic in most modern conflict-driven clause-learning (CDCL) SAT solvers. The precise reason as to why and how restarts enable CDCL solvers to scale efficiently remains obscure. In this paper we address this question, and provide some answers that enabled us to design a new effective machine learning-based restart policy. Specifically, we provide evidence that restarts improve the quality of learnt clauses as measured by one of best known clause quality metrics, namely, literal block distance (LBD). More precisely, we show that more frequent restarts decrease the LBD of learnt clauses, which in turn improves solver performance. We also note that too many restarts can be harmful because of the computational overhead of rebuilding the search tree from scratch too frequently. With this trade-off in mind, between that of learning better clauses vs. the computational overhead of rebuilding the search tree, we introduce a new machine learning-based restart policy that predicts the quality of the next learnt clause based on the history of previously learnt clauses. The restart policy erases the solver's search tree during its run, if it predicts that the quality of the next learnt clause is below some dynamic threshold that is determined by the solver's history on the given input. Our machine learning-based restart policy is based on two observations gleaned from our study of LBDs of learnt clauses. First, we discover that high LBD percentiles can be approximated with z-scores of the normal distribution. Second, we find that LBDs, viewed as a sequence, are correlated and hence the LBDs of past learnt clauses can be used to predict the LBD of future ones. With these observations in place, and techniques to exploit them, our new restart policy is shown to be effective over a large benchmark from the SAT Competition 2014 to 2017.
引用
收藏
页码:94 / 110
页数:17
相关论文
共 50 条
  • [31] Machine Learning-Based Volume Diagnosis
    Wang, Seongmoon
    Wei, Wenlong
    DATE: 2009 DESIGN, AUTOMATION & TEST IN EUROPE CONFERENCE & EXHIBITION, VOLS 1-3, 2009, : 902 - 905
  • [32] A Machine Learning-Based Evaluation Method for Machine Translation
    Kotani, Katsunori
    Yoshimi, Takehiko
    ARTIFICIAL INTELLIGENCE: THEORIES, MODELS AND APPLICATIONS, PROCEEDINGS, 2010, 6040 : 351 - +
  • [33] Extending Clause Learning of SAT Solvers with Boolean Grobner Bases
    Zengler, Christoph
    Kuechlin, Wolfgang
    COMPUTER ALGEBRA IN SCIENTIFIC COMPUTING, 2010, 6244 : 293 - 302
  • [34] On the power of clause-learning SAT solvers as resolution engines
    Pipatsrisawat, Knot
    Darwiche, Adnan
    ARTIFICIAL INTELLIGENCE, 2011, 175 (02) : 512 - 525
  • [36] Extending Clause Learning SAT Solvers with Complete Parity Reasoning
    Laitinen, Tero
    Junttila, Tommi
    Niemela, Ilkka
    2012 IEEE 24TH INTERNATIONAL CONFERENCE ON TOOLS WITH ARTIFICIAL INTELLIGENCE (ICTAI 2012), VOL 1, 2012, : 65 - 72
  • [37] FPGA based accelerator for 3-SAT conflict analysis in SAT solvers
    Safar, M
    El-Kharashi, MW
    Salem, A
    CORRECT HARDWARE DESIGN AND VERIFICATION METHODS, PROCEEDINGS, 2005, 3725 : 384 - 387
  • [38] Coverage-Based Clause Reduction Heuristics for CDCL Solvers
    Nabeshima, Hidetomo
    Inoue, Katsumi
    THEORY AND APPLICATIONS OF SATISFIABILITY TESTING (SAT 2017), 2017, 10491 : 136 - 144
  • [39] Symbolic reachability analysis based on SAT-solvers
    Abdulla, PA
    Bjesse, P
    Eén, N
    TOOLS AND ALGORITHMS FOR THE CONSTRUCTION AND ANALYSIS OF SYSTEMS, 2000, 1785 : 411 - 425
  • [40] NeuroGIFT: Using a Machine Learning Based Sat Solver for Cryptanalysis
    Sun, Ling
    Gerault, David
    Benamira, Adrien
    Peyrin, Thomas
    CYBER SECURITY CRYPTOGRAPHY AND MACHINE LEARNING (CSCML 2020), 2020, 12161 : 62 - 84