Optimization with Objective Functions Learned Through Machine Learning

Title：Optimization with Objective Functions Learned Through Machine Learning

Speaker：Teng Huang (School of Business, University of Connecticut, Ph.D. Candidate)

Host：WANG Shan (Lingnan College, Sun Yat-sen University, Assistant Professor)

Date and Time： April 30th, 2020 (9:00-10:30 a.m.)

Location： Zoom ID: 664 3603 3134   https://cernet.zoom.com.cn/j/66436033134

Language： English+Chinese

Abstract:

My research explores how commonly used machine learning models can be embedded in an optimization problem. In this talk we discuss a novel data-driven decision-making pipeline for solving optimization problems where the objective function is modeled through machine learning. We solve such a decision-making problem in cases when the values of features of a machine learning model depend on the values of decision variables in an optimization model. We explore two real-world applications of this framework in the context of retail network expansion and team formation. In addition, we describe JANOS, an open-source modeling framework that allows users to incorporate linear regression, logistic regression, and neural networks into decision models.

Bio:

Teng Huang is a PhD candidate in the Department of Operations and Information Management at the University of Connecticut School of Business. Her research focuses on integrating machine learning and optimization to facilitate large-scale automated decision-making, and also on quantum computing. She has a paper published at Production and Operations Management and manuscripts under review at INFORMS Journal on Computing and INFORMS Journal on Optimization.