Online Learning and Decision Making with High-Dimensional Data

Individual-level data have become increasingly accessible in the internet era, and the growing availability of user-specific data, such as demographics, geographic, medical records and searching/browsing history, provides decision-makers with unprecedented opportunities to tailor decisions to individual users. Yet, decision-makers’ abilities to use all available information and predict users’ utilities and choices are often impaired by limited samples and/or high computational costs. In this talk, we will apply various state-of-the-art machine-learning algorithms, such as Lasso, MCP and random projection, to tackle these challenges and demonstrate their effectiveness in real-life applications, such as the recommendation system and the assortment optimization system.

Mike Wei is an associate professor of operations management and strategy in the UB School of Management. His research focuses on supply chain management, dynamic pricing, strategic consumer behavior, and high-dimensional machine learning. Wei's work has been published in such leading operations management journals and machine learning conferences as Management Science, Productions and Operations Management, and ICML. He has consulted the in consumer electronics, financial and IT industries.