活动

“Mathematical Programming Computation” Seminar Series: Dr. Xijun Li

Organizers: Shenzhen Research Institute of Big Data (深圳市大数据研究院 in Chinese)

               Shenzhen International Center for Industrial and Applied Mathematics (深圳国际工业与应用数学中心 in Chinese)

Speaker: Dr. Xijun Li

Language: English

Talk title: Order matters: Boosting Mathematical Solver via Machine Learning Techniques

Talk abstract: The trend of using machine learning techniques to improve the mathematical programming solvers has recently drawn lots of attention. In this talk, we will first restate an easily overlooked problem that the order determination leads to performance variability of solvers, which was proposed in decades ago but was not solved well. Then we will present a general solution that utilizes machine learning techniques to find better order for key decision-making component such as reformulation, cut selection, presolving, etc. in mathematical solvers, which boosts the performance of solvers by a large margin. We hope this talk can inspire the future research to better exploit the performance variability via machine learning to improve the solvers.

Bio: Xijun is a senior researcher of HUAWEI Noah’s Ark Lab. Before that, he has received M.Sc. degree from Shanghai Jiao Tong University, P.R. China, in 2018. He is working towards his Ph.D. degree in the University of Science and Technology of China (HUAWEI-USTC Joint Ph.D. Program) under the supervision of Prof. Jie Wang. He has published 10+ papers on top peer-reviewed conferences and journals (such as ICLR, KDD, ICDE, SIGMOD, DAC, TCYB, TAES, etc.) and applied/published 12 patents with Noah's Ark Lab. And he has won the championship of student learderboard in Dual Track of NeurIPS’21 ML4CO Competition. His recent research interests focus on Mathematical Programming Solver, Learning to Optimization (L2O) and Machine Learning for Computer System (ML4CS).

Date & Time: 10:00-11:00 AM, Mar. 6^th, 2023 (Beijing time)

Location: DY103, The Chinese University of Hong Kong, Shenzhen

Zoom Meeting ID: 956 2834 8316 (Passcode: 026260)

Video Link：Order matters: Boosting Mathematical Solver via Machine Learning Techniques

Slide:Order matters: Boosting Mathematical Solver via Machine Learning Techniques