Seminar Vol. 197

Title: Many Inequalities Selection via Machine Learning

Speaker: Ye Luo, The University of Hong Kong

Time: November 15th, 2019  13:30-15:00

Venue: Conference Room 106B, IESR, Zhonghui Building (College of Economics)

About the speaker:

Dr. Ye Luo received his Ph.D from Masschusetts Institute of Technology in 2015. He received his B.S. degree from Massachusetts Institute of Technology in 2010, majored in Mathematics and Economics. Before joining FBE of HKU, he worked as an Assistant Professor at the Economics Department of the University of Florida. Dr. Ye Luo's main research interests include high dimensional econometrics/statistics, machine learning and its empirical applications in economics and finance, for example, applying AI algorithms to develop smart, adaptive automated trading systems, applying big data methods/machine learning in default risk prediction, dynamic demand prediction, etc. He also has interest and expertise in natural language processing.

Dr. Ye Luo has research papers published/forthcoming at Econometrica, Journal of the Royal Statistical Society: Series B, American Economic Review, P&P, etc. Beyond Dr. Ye Luo's academic research, he has a strong interest in connecting the research in data science to the industry. He has given/being invited to give lectures at DiDi, ShunFeng Express, Novartis, etc.

Abstract:

Many economics and operations research problems generate a large set of linear inequalities constraints. The large number of constraints can make optimization and inference problem time costly. We propose a machine learning method, similar to the Dantzig Selector, based on relaxation of Farkas lemma. We prove the asymptotic properties of our selection method, and demonstrate the effectiveness of such selection process by simulation, compare to other more standard methods.