papers AI Learner
The Github is limit! Click to go to the new site.

A Learning based Branch and Bound for Maximum Common Subgraph Problems

2019-05-15
Yan-li Liu, Chu-min Li, Hua Jiang, Kun He

Abstract

Branch-and-bound (BnB) algorithms are widely used to solve combinatorial problems, and the performance crucially depends on its branching heuristic.In this work, we consider a typical problem of maximum common subgraph (MCS), and propose a branching heuristic inspired from reinforcement learning with a goal of reaching a tree leaf as early as possible to greatly reduce the search tree size.Extensive experiments show that our method is beneficial and outperforms current best BnB algorithm for the MCS.

Abstract (translated by Google)
URL

http://arxiv.org/abs/1905.05840

PDF

http://arxiv.org/pdf/1905.05840


Similar Posts

Comments