{"type":"ScholarlyArticle","title":"Learning to Run Heuristics in Tree Search","authors":[{"name":"Elias B. Khalil","slug":"elias-b-khalil","given":"Elias B.","family":"Khalil","affiliation":"School of Computational Science & Engineering , Georgia Institute of Technology , Atlanta , GA , USA"},{"name":"Bistra Dilkina","slug":"bistra-dilkina","given":"Bistra","family":"Dilkina","affiliation":"School of Computational Science & Engineering , Georgia Institute of Technology , Atlanta , GA , USA"},{"name":"George L. Nemhauser","slug":"george-l-nemhauser","given":"George L.","family":"Nemhauser","affiliation":"School of Industrial & Systems Engineering , Georgia Institute of Technology , Atlanta , GA , USA"},{"name":"Shabbir Ahmed","slug":"shabbir-ahmed","given":"Shabbir","family":"Ahmed","affiliation":"School of Industrial & Systems Engineering , Georgia Institute of Technology , Atlanta , GA , USA"},{"name":"Yufen Shao","slug":"yufen-shao","given":"Yufen","family":"Shao","affiliation":"ExxonMobil Upstream Research Company , Houston , TX , USA"}],"abstract":"``Primal heuristics'' are a key contributor to the improved performance of exact branch-and-bound solvers for combinatorial optimization and integer programming. \r\nPerhaps the most crucial question concerning primal heuristics is that of at which nodes they should run, to which the typical answer is via hard-coded rules or fixed solver parameters tuned, offline, by trial-and-error.\r\nAlternatively, a heuristic should be run when it is most likely to succeed, based on the problem instance's characteristics, the state of the search, etc. In this work, we study the problem of deciding at which node a heuristic should be run, such that the overall (primal) performance of the solver is optimized. To our knowledge, this is the first attempt at formalizing and systematically addressing this problem. Central to our approach is the use of Machine Learning (ML) for predicting whether a heuristic will succeed at a given node. We give a theoretical framework for analyzing this decision-making process in a simplified setting, propose a ML approach for modeling heuristic success likelihood, and design practical rules that leverage the ML models to dynamically decide whether to run a heuristic at each node of the search tree. Experimentally, our approach improves the primal performance of a state-of-the-art Mixed Integer Programming solver by up to 6% on a set of benchmark instances, and by up to 60% on a family of hard Independent Set instances.","doi":"10.24963/ijcai.2017/92","url":"https://ijcai.org/proceedings/2017/92","pdf":"https://ijcai.org/proceedings/2017/0092.pdf","markdown":"https://ijcai.org/proceedings/2017/92.md","bibtex":"https://ijcai.org/proceedings/2017/bibtex/92","volume":{"slug":"2017","title":"Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2017","url":"https://ijcai.org/proceedings/2017/","year":2017,"location":"Melbourne, Australia","dates":"19-25 August 2017","editors":"Carles Sierra","isbn":"978-0-9992411-0-3","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2017"},"section":"Constraints and Satisfiability","track":"Main track","number":92,"pages":{"first":659,"last":666},"keywords":["Combinatorial & Heuristic Search: Combinatorial search/optimisation","Constraints and Satisfiability: Constraint Optimisation","Constraints and Satisfiability: Constraints and Satisfiability","Constraints and Satisfiability: Constraint Satisfaction"],"videos":[],"references":[{"position":1,"raw":"[Abraham et al., 2007] David J Abraham, Avrim Blum, and Tuo- mas Sandholm. Clearing algorithms for barter exchange markets: Enabling nationwide kidney exchanges. In Proceedings of the 8th ACM conference on Electronic commerce, pages 295-304. ACM, 2007.","title":"Clearing algorithms for barter exchange markets: Enabling nationwide kidney exchanges","authors":"Avrim David J Abraham, Tuomas Blum, Sandholm","year":"2007","venue":"Proceedings of the 8th ACM conference on Electronic commerce","doi":null,"url":null,"internal_url":null},{"position":2,"raw":"[Achterberg and Wunderling, 2013] Tobias Achterberg and Roland Wunderling. Mixed Integer Programming: Analyzing 12 Years of Progress. In Michael Jünger and Gerhard Reinelt, editors, Facets of Combinatorial Optimization. Springer, 2013.","title":"Mixed Integer Programming: Analyzing 12 Years of Progress","authors":"Wunderling Achterberg, Tobias Achterberg, Roland Wunderling","year":"2013","venue":"Facets of Combinatorial Optimization","doi":null,"url":null,"internal_url":null},{"position":3,"raw":"[Achterberg et al., 2012] Tobias Achterberg, Timo Berthold, and Gregor Hendel. Rounding and propagation heuristics for mixed integer programming. In Operations Research Proceedings 2011, pages 71-76. Springer, 2012.","title":"Rounding and propagation heuristics for mixed integer programming","authors":"Achterberg","year":"2011","venue":"Operations Research Proceedings","doi":null,"url":null,"internal_url":null},{"position":4,"raw":"[Albers, 2003] Susanne Albers. Online algorithms: a survey. Math- ematical Programming, 97(1-2):3-26, 2003.","title":"Online algorithms: a survey","authors":"Susanne Albers","year":"2003","venue":"Mathematical Programming","doi":null,"url":null,"internal_url":null},{"position":5,"raw":"[Andrychowicz et al., 2016] Marcin Andrychowicz, Misha Denil, Sergio Gomez, Matthew W Hoffman, David Pfau, Tom Schaul, and Nando de Freitas. Learning to learn by gradient descent by gradient descent. In Advances in Neural Information Processing Systems, pages 3981-3989, 2016.","title":"Learning to learn by gradient descent by gradient descent","authors":"Marcin Andrychowicz, Misha Denil, Sergio Gomez, Matthew W Hoffman, David Pfau, Tom Schaul, Nando De Freitas","year":"2016","venue":"Advances in Neural Information Processing Systems","doi":null,"url":null,"internal_url":null},{"position":6,"raw":"[Barnhart et al., 1998] Cynthia Barnhart, Natashia L Boland, Lloyd W Clarke, Ellis L Johnson, George L Nemhauser, and Rajesh G Shenoi. Flight string models for aircraft fleeting and routing. Transportation science, 32(3):208-220, 1998.","title":"Flight string models for aircraft fleeting and routing","authors":"Cynthia Barnhart, Natashia L Boland, Lloyd W Clarke, Ellis L Johnson, George L Nemhauser, Rajesh G Shenoi","year":"1998","venue":"Transportation science","doi":null,"url":null,"internal_url":null},{"position":7,"raw":"[Berthold, 2006] Timo Berthold. Primal Heuristics for Mixed In- teger Programs. Master's thesis, Technische Universität Berlin, 2006.","title":"Primal Heuristics for Mixed Integer Programs","authors":"Timo Berthold","year":"2006","venue":null,"doi":null,"url":null,"internal_url":null},{"position":8,"raw":"[Colombi et al., 2016] Marco Colombi, Renata Mansini, and Mar- tin Savelsbergh. The generalized independent set problem: Poly- hedral analysis and solution approaches. European Journal of Operational Research, 2016.","title":"The generalized independent set problem: Polyhedral analysis and solution approaches","authors":"Marco Colombi, Renata Mansini, Martin Savelsbergh","year":"2016","venue":"European Journal of Operational Research","doi":null,"url":null,"internal_url":null},{"position":9,"raw":"[Dilkina and Gomes, 2010] Bistra Dilkina and Carla P Gomes. Solving connected subgraph problems in wildlife conservation. In International Conference on Integration of Artificial Intelli- gence (AI) and Operations Research (OR) Techniques in Con- straint Programming, pages 102-116. Springer, 2010.","title":"Solving connected subgraph problems in wildlife conservation","authors":"Gomes Dilkina, Bistra Dilkina, Carla P Gomes","year":"2010","venue":"International Conference on Integration of Artificial Intelligence (AI) and Operations Research (OR) Techniques in Constraint Programming","doi":null,"url":null,"internal_url":null},{"position":10,"raw":"[Gamrath et al., 2016] Gerald Gamrath, Tobias Fischer, Tristan Gally, Ambros M. Gleixner, Gregor Hendel, Thorsten Koch, Stephen J. Maher, Matthias Miltenberger, Benjamin Müller, Marc E. Pfetsch, Christian Puchert, Daniel Rehfeldt, Sebastian Schenker, Robert Schwarz, Felipe Serrano, Yuji Shinano, Stefan Vigerske, Dieter Weninger, Michael Winkler, Jonas T. Witt, and Jakob Witzig. The SCIP Optimization Suite 3.2. Technical Re- port 15-60, ZIB, Takustr.7, 14195 Berlin, 2016.","title":"The SCIP Optimization Suite 3.2","authors":"Gerald Gamrath, Tobias Fischer, Tristan Gally, Ambros M Gleixner, Gregor Hendel, Thorsten Koch, Stephen J Maher, Matthias Miltenberger, Benjamin Müller, Marc E Pfetsch, Christian Puchert, Daniel Rehfeldt, Sebastian Schenker, Robert Schwarz, Felipe Serrano, Yuji Shinano, Stefan Vigerske, Dieter Weninger, Michael Winkler, Jonas T Witt, Jakob Witzig","year":"2016","venue":null,"doi":null,"url":null,"internal_url":null},{"position":11,"raw":"[He and Garcia, 2009] Haibo He and Edwardo A Garcia. Learn- ing from imbalanced data. IEEE Transactions on knowledge and data engineering, 21(9):1263-1284, 2009.","title":"Learning from imbalanced data","authors":"Haibo He, Edwardo A Garcia","year":"2009","venue":"IEEE Transactions on knowledge and data engineering","doi":null,"url":null,"internal_url":null},{"position":12,"raw":"[Hendel, 2011] Gregor Hendel. New rounding and propagation heuristics for mixed integer programming, 2011.","title":"New rounding and propagation heuristics for mixed integer programming","authors":"Gregor Hendel","year":"2011","venue":null,"doi":null,"url":null,"internal_url":null},{"position":13,"raw":"[Hochbaum and Pathria, 1997] Dorit S Hochbaum and Anu Pathria. Forest harvesting and minimum cuts: a new approach to handling spatial constraints. Forest Science, 43(4):544-554, 1997.","title":"Forest harvesting and minimum cuts: a new approach to handling spatial constraints","authors":"S Dorit, Anu Hochbaum, Pathria","year":"1997","venue":"Forest Science","doi":null,"url":null,"internal_url":null},{"position":14,"raw":"[Hutter et al., 2009] Frank Hutter, Holger H Hoos, Kevin Leyton- Brown, and Thomas Stützle. ParamILS: an automatic algorithm configuration framework. JAIR, 36(1):267-306, 2009.","title":"ParamILS: an automatic algorithm configuration framework","authors":"Frank Hutter, H Holger, Kevin Hoos, Thomas Leyton-Brown, Stützle","year":"2009","venue":"JAIR","doi":null,"url":null,"internal_url":null},{"position":15,"raw":"[IBM, 2014] IBM. CPLEX User's Manual, Version 12.6.1, 2014.","title":"IBM. CPLEX User's Manual","authors":null,"year":"2014","venue":null,"doi":null,"url":null,"internal_url":null},{"position":16,"raw":"[Johnson and Trick, 1996] David S Johnson and Michael A Trick. Cliques, coloring, and satisfiability: second DIMACS implemen- tation challenge, October 11-13, 1993, volume 26. American Mathematical Soc., 1996.","title":"Cliques, coloring, and satisfiability: second DIMACS implementation challenge","authors":"S David, Michael A Johnson, Trick","year":"1993","venue":null,"doi":null,"url":null,"internal_url":null},{"position":17,"raw":"[Khalil et al., 2016] Elias Boutros Khalil, Pierre Le Bodic, Le Song, George L Nemhauser, and Bistra N Dilkina. Learn- ing to branch in mixed integer programming. In AAAI, pages 724-731, 2016.","title":"Learning to branch in mixed integer programming","authors":"Elias Boutros Khalil, Pierre Le Bodic, Le Song, George L Nemhauser, Bistra N Dilkina","year":"2016","venue":"AAAI","doi":null,"url":null,"internal_url":null},{"position":18,"raw":"[Koch et al., 2011] Thorsten Koch, Tobias Achterberg, Erling An- dersen, Oliver Bastert, Timo Berthold, Robert E Bixby, Emi- lie Danna, Gerald Gamrath, Ambros M Gleixner, Stefan Heinz, et al. MIPLIB 2010. Mathematical Programming Computation, 3(2):103-163, 2011.","title":"Mathematical Programming Computation","authors":"Thorsten Koch, Tobias Achterberg, Erling Andersen, Oliver Bastert, Timo Berthold, Robert E Bixby, Emilie Danna, Gerald Gamrath, Ambros M Gleixner, Stefan Heinz","year":"2010","venue":"Mathematical Programming Computation","doi":null,"url":null,"internal_url":null},{"position":19,"raw":"[Kruber et al., 2016] Markus Kruber, Marco E Lübbecke, and Axel Parmentier. Learning when to use a decomposition. 2016.","title":"Learning when to use a decomposition","authors":"Markus Kruber, Marco E Lübbecke, Axel Parmentier","year":"2016","venue":null,"doi":null,"url":null,"internal_url":null},{"position":20,"raw":"[Lee et al., 1999] Eva K Lee, Richard J Gallagher, David Silvern, Cheng-Shie Wuu, and Marco Zaider. Treatment planning for brachytherapy: an integer programming model, two computa- tional approaches and experiments with permanent prostate im- plant planning. Physics in Medicine and Biology, 44(1):145, 1999.","title":"Treatment planning for brachytherapy: an integer programming model, two computational approaches and experiments with permanent prostate implant planning","authors":"Richard J Eva K Lee, David Gallagher, Cheng-Shie Silvern, Marco Wuu, Zaider","year":"1999","venue":"Physics in Medicine and Biology","doi":null,"url":null,"internal_url":null},{"position":21,"raw":"[Lodi and Tramontani, 2013] Andrea Lodi and Andrea Tramontani. Performance variability in mixed-integer programming. Tutori- als in Operations Research: Theory Driven by Influential Appli- cations, pages 1-12, 2013.","title":"Performance variability in mixed-integer programming","authors":"Andrea Lodi, Andrea Tramontani","year":"2013","venue":"Tutorials in Operations Research: Theory Driven by Influential Applications","doi":null,"url":null,"internal_url":null},{"position":22,"raw":"[Nemhauser and Trick, 1998] George L Nemhauser and Michael A Trick. Scheduling a major college basketball conference. Oper- ations Research, 46(1):1-8, 1998.","title":"Scheduling a major college basketball conference","authors":"L George, Michael A Nemhauser, Trick","year":"1998","venue":"Operations Research","doi":null,"url":null,"internal_url":null},{"position":23,"raw":"[Papageorgiou et al., 2014] Dimitri J Papageorgiou, George L Nemhauser, Joel Sokol, Myun-Seok Cheon, and Ahmet B Keha. MIRPLib -A library of maritime inventory routing problem in- stances: Survey, core model, and benchmark results. European Journal of Operational Research, 235(2):350-366, 2014.","title":"MIRPLib -A library of maritime inventory routing problem instances: Survey, core model, and benchmark results","authors":"Dimitri J Papageorgiou, George L Nemhauser, Joel Sokol, Myun-Seok Cheon, Ahmet B Keha","year":"2014","venue":"European Journal of Operational Research","doi":null,"url":null,"internal_url":null},{"position":24,"raw":"[Pedregosa et al., 2011] F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. Scikit-learn: Ma- chine learning in Python. Journal of Machine Learning Research, 12:2825-2830, 2011.","title":"Scikit-learn: Machine learning in Python","authors":"F Pedregosa, G Varoquaux, A Gramfort, V Michel, B Thirion, O Grisel, M Blondel, P Prettenhofer, R Weiss, V Dubourg, J Vanderplas, A Passos, D Cournapeau, M Brucher, M Perrot, E Duchesnay","year":"2011","venue":"Journal of Machine Learning Research","doi":null,"url":null,"internal_url":null},{"position":25,"raw":"[Sabharwal et al., 2012] Ashish Sabharwal, Horst Samulowitz, and Chandra Reddy. Guiding combinatorial optimization with uct. In International Conference on Integration of Artificial Intelli- gence (AI) and Operations Research (OR) Techniques in Con- straint Programming, pages 356-361. Springer, 2012.","title":"Guiding combinatorial optimization with uct","authors":"Ashish Sabharwal, Horst Samulowitz, Chandra Reddy","year":"2012","venue":"International Conference on Integration of Artificial Intelligence (AI) and Operations Research (OR) Techniques in Constraint Programming","doi":null,"url":null,"internal_url":null},{"position":26,"raw":"[Samulowitz and Memisevic, 2007] Horst Samulowitz and Roland Memisevic. Learning to solve QBF. In AAAI, 2007.","title":"Learning to solve QBF","authors":"Memisevic Samulowitz, Horst Samulowitz, Roland Memisevic","year":"2007","venue":null,"doi":null,"url":null,"internal_url":null},{"position":27,"raw":"[Zhang and Dietterich, 1995] Wei Zhang and Thomas G Dietterich. A reinforcement learning approach to job-shop scheduling. In IJCAI, volume 95, pages 1114-1120. Citeseer, 1995.","title":"A reinforcement learning approach to job-shop scheduling","authors":"Wei Zhang, Thomas G Dietterich","year":"1995","venue":"IJCAI","doi":null,"url":null,"internal_url":"/proceedings/1995/144"}],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}