Learning to Solve QBF

We present a novel approach to solving Quantified Boolean Formulas (QBF) that combines a search-based QBF solver with marhine learning techniques. We show how classification methods can be used to predict run-times and to choose optimal heuristics both within a portfolio-based, and within a dynamic, online approach. In the dynamic method variables are set to a truth value according to a scheme that tries to maximize the probability of successfully solving the remaining sub-problem efficiently. Since each variable assignment can drastically change the problem-structure, new heuristics are chosen dynamically, and a classifier is used online to predict the usefulness of each heuristic. Experimental results on a large corpus of example problems show the usefulness of our approach in terms of run-time as well as the ability to solve previously unsolved problem instances.

[1]  R. Memisevic.  An introduction to structured discriminative learning , 2006 .

[2]  Kevin Leyton-Brown,et al.  Performance Prediction and Automated Tuning of Randomized and Parametric Algorithms , 2006, CP.

[3]  Fahiem Bacchus,et al.  Binary Clause Reasoning in QBF , 2006, SAT.

[4]  Rolf Drechsler,et al.  Post-verification debugging of hierarchical designs , 2005, ICCAD-2005. IEEE/ACM International Conference on Computer-Aided Design, 2005..

[5]  Reinhold Letz,et al.  Lemma and Model Caching in Decision Procedures for Quantified Boolean Formulas , 2002, TABLEAUX.

[6]  Fahiem Bacchus,et al.  Dynamically Partitioning for Solving QBF , 2007, SAT.

[7]  Michail G. Lagoudakis,et al.  Learning to Select Branching Rules in the DPLL Procedure for Satisfiability , 2001, Electron. Notes Discret. Math..

[8]  S. Malik,et al.  Towards symmetric treatment of con?icts and satisfaction in quanti-fied Boolean satisfiability solv , 2002 .

[9]  Eric R. Ziegel,et al.  The Elements of Statistical Learning , 2003, Technometrics.

[10]  Sharad Malik,et al.  Chaff: engineering an efficient SAT solver , 2001, Proceedings of the 38th Design Automation Conference (IEEE Cat. No.01CH37232).

[11]  Armin Biere,et al.  Resolve and Expand , 2004, SAT.

[12]  Luca Pulina,et al.  Report of the Third QBF Solvers Evaluation , 2006, J. Satisf. Boolean Model. Comput..

[13]  Y. Shoham,et al.  SATzilla : An Algorithm Portfolio for SAT ∗ , 2004 .

[14]  Donald W. Loveland,et al.  A machine program for theorem-proving , 2011, CACM.