Abstract: In this talk, I will introduce AI-based combinatorial optimization algorithms designed to efficiently solve various combinatorial optimization problems. AI-based combinatorial optimization techniques have diverse industrial applications, such as the development of new chemicals and drugs, optimization of logistics and transportation systems, and semiconductor chip design. First, I will present the structure and training methods of AI models for solving combinatorial optimization problems. Furthermore, I will discuss the development of the Neural-Guided Improvement Search technique and the construction of an end-to-end learning framework that enhances the performance of combinatorial optimization solvers by integrating solution sampling and search processes. Finally, I will present case studies that demonstrate the application of these techniques to practical problems.

Bio: Professor Jinkyoo Park is currently an Associate Professor in the Department of Industrial and Systems Engineering at KAIST and an affiliated professor at the Kim Jaechul Graduate School of AI. He received his bachelor's degree in Architecture from Seoul National University in 2009 and his Ph.D. from Stanford University in 2016. Professor Park is also the CEO of Omelet, a startup that provides AI-based combinatorial optimization solvers to the industry.

Abstract:  Neural Combinatorial Optimization (NCO) aims to learn directly from data a neural network that can solve complex combinatorial optimization problems like the travel salesman problem (TSP) and the Capacitated Vehicle Routing Problem (CVRP). Existing NCO models achieve good performance on small-scale problem instances but fail to generalize to solve large-scale ones. In this talk, I will systematically review these existing NCO methods and introduce their basic principles. Thereafter, I will reveal some possible reasons for their poor large-scale generalization capability. Beyond that, I will introduce a new model structure, Light Encoder and Heavy Decoder (LEHD), and illustrate why it can achieve excellent large-scale generalization performance. A data-efficient training scheme and a flexible solution construction mechanism will be presented along with the LEHD model. Finally, some experimental results on TSP and CVRP will be provided to indicate the superiority of the LEHD.

Bio:  Dr. Zhenkun Wang received his Ph.D. degree from Xidian University in 2016. From 2017 to 2020, he was a post-doctoral research fellow at Nanyang Technological University and the City University of Hong Kong in Singapore and Hong Kong, respectively. In June 2020, he joined the Southern University of Science and Technology as an Assistant Professor. His research interests mainly include multiobjective optimization, combinatorial optimization, evolutionary computation, and deep learning. Dr. Zhenkun Wang has published over 50 academic papers in leading journals (e.g., IEEE TEVC, IEEE TCYB, IEEE TSMCA, and IEEE TITS) and top AI conferences  (e.g., ICML, NeurIPS, AAAI, IJCAI, and KDD), and has won two provincial and ministerial science and technology awards. He is an IEEE senior member, serves as associate editor of three journals, and is the chair of student activities in the IEEE CIS Shenzhen Chapter. Moreover, he has served as the Competition Chair of EMO 2021 and the PC Member of several top conferences.

Abstract:  Recent advances in machine learning and deep learning for combinatorial optimization have significantly impacted the fields of computational intelligence and operational research. In this talk, I will present the challenges faced by traditional algorithms and deep reinforcement learning algorithms in solving combinatorial optimization problems. Then, I will showcase our recent progress in developing hybrid algorithms that leverage machine learning and reinforcement learning techniques. This talk will end by exploring potential future directions for this exciting and rapidly evolving area.

Bio:  Dr. Yan Jin is an associate professor in the School of Computer Science and Technology, Huazhong University of Science and Technology (HUST), China. She received the Ph.D. degree in Computer Science from University of Angers, France, in 2016. She was a StarTrack Visiting Faculty at Microsoft Research Asia in 2021. Her research focuses on the design of effective algorithms for solving combinatorial optimization problems and practical applications, including reinforcement learning, machine learning based search approaches and hybrid evolutionary algorithms. She has published over 30 papers and served as a PC Member for conferences such as IJCAI, AAAI and AAMAS, as well as a reviewer for several well-known journals.