Please use this identifier to cite or link to this item: https://doi.org/10.1016/j.knosys.2020.106244
Title: A reinforcement learning approach for optimizing multiple traveling salesman problems over graphs
Authors: YUJIAO HU
YUAN YAO
LEE WEE SUN 
Keywords: multi-agent reinforcement learning
combinatorial optimization problems
multiple traveling salesman problems
graph neural networks
policy networks
Issue Date: 27-Sep-2020
Publisher: Elsevier B.V.
Citation: YUJIAO HU, YUAN YAO, LEE WEE SUN (2020-09-27). A reinforcement learning approach for optimizing multiple traveling salesman problems over graphs. Kowledge-Based Systems 204 : 106244. ScholarBank@NUS Repository. https://doi.org/10.1016/j.knosys.2020.106244
Abstract: This paper proposes a learning-based approach to optimize the multiple traveling salesman problem (MTSP), which is one classic representative of cooperative combinatorial optimization problems. The MTSP is interesting to study, because the problem arises from numerous practical applications and efficient approaches to optimize the MTSP can potentially be adapted for other cooperative optimization problems. However, the MTSP is rarely researched in the deep learning domain because of certain difficulties, including the huge search space, the lack of training data that is labeled with optimal solutions and the lack of architectures that extract interactive behaviors among agents. This paper constructs an architecture consisting of a shared graph neural network and distributed policy networks to learn a common policy representation to produce near-optimal solutions for the MTSP. We use a reinforcement learning approach to train the model, overcoming the requirement data labeled with ground truth. We use a two-stage approach, where reinforcement learning is used to learn an allocation of agents to vertices, and a regular optimization method is used to solve the single-agent traveling salesman problems associated with each agent. We introduce a S-samples batch training method to reduce the variance of the gradient, improving the performance significantly. Experiments demonstrate our approach successfully learns a strong policy representation that outperforms integer linear programming and heuristic algorithms, especially on large scale problems.
Source Title: Kowledge-Based Systems
URI: https://scholarbank.nus.edu.sg/handle/10635/177387
ISSN: 09507051
DOI: 10.1016/j.knosys.2020.106244
Appears in Collections:Staff Publications
Elements

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
MinMaxMTSPpreprint.pdf1.9 MBAdobe PDF

OPEN

Pre-printView/Download

Google ScholarTM

Check

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.