Generative Adversarial Imitation Learning to Search in Branch-and-Bound Algorithms

被引:2
作者
Wang, Qi [1 ]
Blackley, Suzanne, V [2 ]
Tang, Chunlei [2 ]
机构
[1] Fudan Univ, Shanghai 200438, Peoples R China
[2] Harvard Med Sch, Boston, MA 02120 USA
来源
DATABASE SYSTEMS FOR ADVANCED APPLICATIONS, DASFAA 2022, PT II | 2022年
关键词
Combinatorial optimization; Reinforcement learning; Branch-and-bound; Generative adversarial imitation learning;
D O I
10.1007/978-3-031-00126-0_51
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Recent studies have shown that reinforcement learning (RL) can provide state-of-the-art performance at learning sophisticated heuristics by exploiting the shared internal structure combinatorial optimization instances in the data. However, existing RL-based methods require too much trial-and-error reliant on sophisticated reward engineering, which is laborious and inefficient for practical applications. This paper proposes a novel framework (RAIL) that combines RL and generative adversarial imitation learning (GAIL) to meet the challenge by searching in branch-and-bound algorithms. RAIL has a policy architecture with dual decoders, corresponding to the sequence decoding of RL and the edge decoding of GAIL, respectively. The two complement each other and restrict each other to improve the learned policy and reward function iteratively.
引用
收藏
页码:673 / 680
页数:8
相关论文
共 50 条
[21]   On estimating workload in interval branch-and-bound global optimization algorithms [J].
Berenguel, Jose L. ;
Casado, L. G. ;
Garcia, I. ;
Hendrix, Eligius M. T. .
JOURNAL OF GLOBAL OPTIMIZATION, 2013, 56 (03) :821-844
[22]   Techniques for accelerating branch-and-bound algorithms dedicated to sparse optimization [J].
Samain, Gwenael ;
Ninin, Jordan ;
Bourguignon, Sebastien .
OPTIMIZATION METHODS & SOFTWARE, 2024, 40 (01) :4-41
[23]   Branch-and-bound algorithms for simple assembly line balancing problem [J].
Liu, S. B. ;
Ng, K. M. ;
Ong, H. L. .
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY, 2008, 36 (1-2) :169-177
[24]   On estimating workload in interval branch-and-bound global optimization algorithms [J].
José L. Berenguel ;
L. G. Casado ;
I. García ;
Eligius M. T. Hendrix .
Journal of Global Optimization, 2013, 56 :821-844
[25]   Multiagent Autonomous Source Search Using Submodularity and Branch-and-Bound [J].
Xu, Xiaoling ;
Marelli, Damian ;
Meng, Wei ;
Cai, Qianqian ;
Fu, Minyue .
UNMANNED SYSTEMS, 2024, 12 (01) :19-28
[26]   Branch-and-Bound Guided Search for Critical Elements in State Estimation [J].
Augusto, Andre Abel ;
Do Coutto Filho, Milton Brown ;
Stacchini de Souza, Julio Cesar ;
Ribeiro Guimaraens, Marcio Andre .
IEEE TRANSACTIONS ON POWER SYSTEMS, 2019, 34 (03) :2292-2301
[27]   A Branch-and-Bound Approach for Tautomer Enumeration [J].
Thalheim, Torsten ;
Wagner, Barbara ;
Kuehne, Ralph ;
Middendorf, Martin ;
Schueuermann, Gerrit .
MOLECULAR INFORMATICS, 2015, 34 (05) :263-275
[28]   Learning Temporal Strategic Relationships using Generative Adversarial Imitation Learning [J].
Fernando, Tharindu ;
Denman, Simon ;
Sridharan, Sridha ;
Fookes, Clinton .
PROCEEDINGS OF THE 17TH INTERNATIONAL CONFERENCE ON AUTONOMOUS AGENTS AND MULTIAGENT SYSTEMS (AAMAS' 18), 2018, :113-121
[29]   Customized Generative Adversarial Imitation Learning for Driving Behavior Modeling in Traffic Simulation [J].
Zhu, Zhongyuan ;
Jiang, Zhuoxuan ;
Zhang, Xuefeng ;
Guo, Jifu ;
Xian, Kai ;
Zhang, Tianyang ;
Ren, Jiawei .
JOURNAL OF ADVANCED TRANSPORTATION, 2025, 2025 (01)
[30]   Generative Adversarial Imitation Learning Based Bicycle Behaviors Simulation on Road Segments [J].
Wei, Shuqiao ;
Ni, Ying ;
Sun, Jian ;
Qiu, Hongtong .
Jiaotong Yunshu Xitong Gongcheng Yu Xinxi/Journal of Transportation Systems Engineering and Information Technology, 2024, 24 (04) :105-115