License: 
Creative Commons Attribution 4.0 International license (CC BY 4.0)
When quoting this document, please refer to the following
DOI: 10.4230/LIPIcs.CP.2023.25
URN: urn:nbn:de:0030-drops-190625
URL: http://dagstuhl.sunsite.rwth-aachen.de/volltexte/2023/19062/
Marty, Tom ;
François, Tristan ;
Tessier, Pierre ;
Gautier, Louis ;
Rousseau, Louis-Martin ;
Cappart, Quentin
Learning a Generic Value-Selection Heuristic Inside a Constraint Programming Solver
Abstract
Constraint programming is known for being an efficient approach to solving combinatorial problems. Important design choices in a solver are the branching heuristics, designed to lead the search to the best solutions in a minimum amount of time. However, developing these heuristics is a time-consuming process that requires problem-specific expertise. This observation has motivated many efforts to use machine learning to automatically learn efficient heuristics without expert intervention. Although several generic variable-selection heuristics are available in the literature, the options for value-selection heuristics are more scarce. We propose to tackle this issue by introducing a generic learning procedure that can be used to obtain a value-selection heuristic inside a constraint programming solver. This has been achieved thanks to the combination of a deep Q-learning algorithm, a tailored reward signal, and a heterogeneous graph neural network. Experiments on graph coloring, maximum independent set, and maximum cut problems show that this framework competes with the well-known impact-based and activity-based search heuristics and can find solutions close to optimality without requiring a large number of backtracks.
BibTeX - Entry
@InProceedings{marty_et_al:LIPIcs.CP.2023.25,
author = {Marty, Tom and Fran\c{c}ois, Tristan and Tessier, Pierre and Gautier, Louis and Rousseau, Louis-Martin and Cappart, Quentin},
title = {{Learning a Generic Value-Selection Heuristic Inside a Constraint Programming Solver}},
booktitle = {29th International Conference on Principles and Practice of Constraint Programming (CP 2023)},
pages = {25:1--25:19},
series = {Leibniz International Proceedings in Informatics (LIPIcs)},
ISBN = {978-3-95977-300-3},
ISSN = {1868-8969},
year = {2023},
volume = {280},
editor = {Yap, Roland H. C.},
publisher = {Schloss Dagstuhl -- Leibniz-Zentrum f{\"u}r Informatik},
address = {Dagstuhl, Germany},
URL = {https://drops.dagstuhl.de/opus/volltexte/2023/19062},
URN = {urn:nbn:de:0030-drops-190625},
doi = {10.4230/LIPIcs.CP.2023.25},
annote = {Keywords: Branching heuristic, Deep reinforcement learning}
}
