Machine Learning for First-Order Theorem Proving Learning to Select a Good Heuristic

被引:52
|
作者
Bridge, James P. [1 ]
Holden, Sean B. [1 ]
Paulson, Lawrence C. [1 ]
机构
[1] Univ Cambridge, Comp Lab, Cambridge CB3 0FD, England
基金
英国工程与自然科学研究理事会;
关键词
Automatic theorem proving; Machine learning; First-order logic with equality; Feature selection; Support vector machines; Gaussian processes; CLASSIFICATION; SAT;
D O I
10.1007/s10817-014-9301-5
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We applied two state-of-the-art machine learning techniques to the problem of selecting a good heuristic in a first-order theorem prover. Our aim was to demonstrate that sufficient information is available from simple feature measurements of a conjecture and axioms to determine a good choice of heuristic, and that the choice process can be automatically learned. Selecting from a set of 5 heuristics, the learned results are better than any single heuristic. The same results are also comparable to the prover's own heuristic selection method, which has access to 82 heuristics including the 5 used by our method, and which required additional human expertise to guide its design. One version of our system is able to decline proof attempts. This achieves a significant reduction in total time required, while at the same time causing only a moderate reduction in the number of theorems proved. To our knowledge no earlier system has had this capability.
引用
收藏
页码:141 / 172
页数:32
相关论文
共 50 条
  • [31] A first-order isomorphism theorem
    Allender, E
    Balcazar, J
    Immerman, N
    SIAM JOURNAL ON COMPUTING, 1997, 26 (02) : 557 - 567
  • [32] Proving Semantic Properties as First-Order Satisfiability
    Lucas, Salvador
    PROCEEDINGS OF THE TWENTY-NINTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2020, : 5075 - 5079
  • [33] Learning probabilities for noisy first-order rules
    Koller, D
    Pfeffer, A
    IJCAI-97 - PROCEEDINGS OF THE FIFTEENTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOLS 1 AND 2, 1997, : 1316 - 1321
  • [34] On the Parameterized Complexity of Learning First-Order Logic
    van Bergerem, Steffen
    Grohe, Martin
    Ritzert, Martin
    PROCEEDINGS OF THE 41ST ACM SIGMOD-SIGACT-SIGAI SYMPOSIUM ON PRINCIPLES OF DATABASE SYSTEMS (PODS '22), 2022, : 337 - 346
  • [35] Distributed Learning Systems with First-Order Methods
    Liu, Ji
    Zhang, Ce
    FOUNDATIONS AND TRENDS IN DATABASES, 2020, 9 (01): : 1 - 100
  • [36] Implicitly Learning to Reason in First-Order Logic
    Belle, Vaishak
    Juba, Brendan
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 32 (NIPS 2019), 2019, 32
  • [37] Holistic Deductive Framework Theorem Proving Based on Standard Contradiction Separation For First-Order Logic
    Cao, Feng
    Xu, Yang
    Zhong, Jian
    Wu, Guanfeng
    2017 12TH INTERNATIONAL CONFERENCE ON INTELLIGENT SYSTEMS AND KNOWLEDGE ENGINEERING (IEEE ISKE), 2017,
  • [38] Reasoning in the OWL 2 Full Ontology Language Using First-Order Automated Theorem Proving
    Schneider, Michael
    Sutcliffe, Geoff
    AUTOMATED DEDUCTION - CADA-23, 2011, 6803 : 461 - 475
  • [39] Application of Multi-Clause Synergized Deduction in First-Order Logic Automated Theorem Proving
    Cao F.
    Xu Y.
    Chen S.
    Wu G.
    Chang W.
    Xinan Jiaotong Daxue Xuebao/Journal of Southwest Jiaotong University, 2020, 55 (02): : 401 - 408and427
  • [40] A Machine Learning Based First-Order Sea Clutter Region Extraction Method for HFSWR
    Li, Yang
    Wang, Xinyang
    Zhang, Ning
    Wang, Wenxing
    Zhang, Qiming
    Ding, Wenbo
    Wu, Longshan
    2019 IEEE INTERNATIONAL SYMPOSIUM ON ANTENNAS AND PROPAGATION AND USNC-URSI RADIO SCIENCE MEETING, 2019, : 173 - 174