Li, Hongbo
[Verfasser:in];
Wu, Yaling
[Verfasser:in];
Yin, Minghao
[Verfasser:in];
Li, Zhanshan
[Verfasser:in]
;
Hongbo Li and Yaling Wu and Minghao Yin and Zhanshan Li
[Mitwirkende:r]
A Portfolio-Based Approach to Select Efficient Variable Ordering Heuristics for Constraint Satisfaction Problems
Anmerkungen:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Beschreibung:
Variable ordering heuristics (VOH) play a central role in solving Constraint Satisfaction Problems (CSP). The performance of different VOHs may vary greatly in solving the same CSP instance. In this paper, we propose an approach to select efficient VOHs for solving different CSP instances. The approach contains two phases. The first phase is a probing procedure that runs a simple portfolio strategy containing several different VOHs. The portfolio tries to use each of the candidate VOHs to guide backtracking search to solve the CSP instance within a limited number of failures. If the CSP is not solved by the portfolio, one of the candidates is selected for it by analysing the information attached in the search trees generated by the candidates. The second phase uses the selected VOH to guide backtracking search to solve the CSP. The experiments are run with the MiniZinc benchmark suite and four different VOHs which are considered as the state of the art are involved. The results show that the proposed approach finds the best VOH for more than 67% instances and it solves more instances than all the candidate VOHs and an adaptive VOH based on Multi-Armed Bandit. It could be an effective adaptive search strategy for black-box CSP solvers.