Guiding Combinatorial Optimization with UCT

Sabharwal, Ashish; Samulowitz, Horst; Reddy, Chandra

doi:10.1007/978-3-642-29828-8_23

Cited by 40 publications

(31 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the study performed by Sabharwal and Samulowitz [179], UCT is applied to guide Mixed Integer Programming (MIP), comparing the performance of the UCT based node selection with that of CPLEX, a traditional MIP solver, and best-first, breadth-first and depth-first strategies, showing very promising results. 35.…”

Section: Mixed Integer Programmingmentioning

confidence: 99%

A Survey of Monte Carlo Tree Search Methods

Browne

Powley

Whitehouse

et al. 2012

IEEE Trans. Comput. Intell. AI Games

2,318

1,535

View full text Add to dashboard Cite

Monte Carlo Tree Search (MCTS) is a recently proposed search method that combines the precision of tree search with the generality of random sampling. It has received considerable interest due to its spectacular success in the difficult problem of computer Go, but has also proved beneficial in a range of other domains. This paper is a survey of the literature to date, intended to provide a snapshot of the state of the art after the first five years of MCTS research. We outline the core algorithm's derivation, impart some structure on the many variations and enhancements that have been proposed, and summarise the results from the key game and non-game domains to which MCTS methods have been applied. A number of open research questions indicate that the field is ripe for future work.

show abstract

Section: Mixed Integer Programmingmentioning

confidence: 99%

A Survey of Monte Carlo Tree Search Methods

Browne

Powley

Whitehouse

et al. 2012

IEEE Trans. Comput. Intell. AI Games

2,318

1,535

View full text Add to dashboard Cite

show abstract

“…A few work have pioneered the use of MCTS to explore a tree-structured assignment search space, in order to solve combinatorial optimization or satisfiability problem instances. In [23], MCTS is applied to Mixed Integer Programming, and used to control the selection of the top nodes in the CPLEX solver; the node reward is set to the maximal value of solutions built on this node. In [24], MCTS is applied to Job Shop Scheduling problems; it is viewed as an alternative to Pilot or roll-out methods, featuring an integrated and smart look-ahead strategy.…”

Section: Monte Carlo Tree Searchmentioning

confidence: 99%

“…This paper advocates the use of another ML approach, namely reinforcement learning (RL) [8], to support the CP search. Taking inspiration from earlier work [25,24,23,9], the paper contribution is to extend the Monte-Carlo Tree Search (MCTS) algorithm to control the exploration of the CP search tree.…”

Section: Introductionmentioning

confidence: 99%

Bandit-Based Search for Constraint Programming

Loth

Sebag

Hamadi

et al. 2013

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Constraint Programming (CP) solvers classically explore the solution space using tree-search based heuristics. Monte-Carlo Tree Search (MCTS), aimed at optimal sequential decision making under uncertainty, gradually grows a search tree to explore the most promising regions according to a specified reward function. At the crossroad of CP and MCTS, this paper presents the Bandit Search for Constraint Programming (BaSCoP) algorithm, adapting MCTS to the specifics of the CP search. This contribution relies on i) a generic reward function suited to CP and compatible with a multiple restart strategy; ii) the use of depthfirst search as roll-out procedure in MCTS. BaSCoP, on the top of the Gecode constraint solver, is shown to significantly improve on depth-first search on some CP benchmark suites, demonstrating its relevance as a generic yet robust CP search method.

show abstract

“…Different node selection strategies are proposed in the Solving Constraint Integer Programs (SCIP) optimization code [22]. Depth-first, breadth-first and best-first search strategies are available.…”

Section: Introductionmentioning

confidence: 99%

“…The score of the visited nodes are decreased in order to diversify the search (i.e., to avoid that the same path is followed every time). A conclusion found in [22] is that UCT is costly but can improve the performances when limited to the first nodes of the search.…”

Section: Introductionmentioning

confidence: 99%

Node selection strategies in interval Branch and Bound algorithms

Neveu¹,

Trombettoni²,

Araya³

2015

J Glob Optim

View full text Add to dashboard Cite

International audienceWe present in this article new strategies for selecting nodes in interval Branch and Bound algorithms for constrained global optimization. For a minimization problem the standard best-first strategy selects a node with the smallest lower bound of the objective function estimate. We first propose new node selection policies where an upper bound of each node/box is also taken into account. The good accuracy of this upper bound achieved by several contracting operators leads to a good performance of the node selection rule based on this criterion. We propose another strategy that also makes a trade-off between diversification and intensification by greedily diving into potential feasible regions at each node of the best-first search. These new strategies obtain better experimental results than classical best-first search on difficult constrained global optimization instances

show abstract

Guiding Combinatorial Optimization with UCT

Cited by 40 publications

References 5 publications

A Survey of Monte Carlo Tree Search Methods

A Survey of Monte Carlo Tree Search Methods

Bandit-Based Search for Constraint Programming

Node selection strategies in interval Branch and Bound algorithms

Contact Info

Product

Resources

About