- Tuesday, March 3, 2020
- 3:30 PM–4:30 PM
- Science Building 010
Learn how the famous "multi-armed bandits" problem, used to model decisions in finance, medicine, and advertising, can be solved by simple reinforcement learning algorithms.
The multi-armed bandit problem has been used for decades to model decisions in domains such as finance, medicine, and internet display advertising. The problem can be approximately solved using simple reinforcement learning algorithms. This talk will introduce the multi-armed bandit problem and reinforcement learning, then demonstrate how two different reinforcement learning algorithms can be used to find well-motivated approximate solutions for the multi-armed bandit problem. The talk will conclude with a description of extensions and variants of the multi-armed bandit problem, other types of reinforcement learning, and general comments on the challenges associated with reinforcement learning relative to other types of machine learning.
Michael Bloem is a Principal Data Scientist at Mosaic Data Science where he leads and executes the design, development, and deployment of data science-enabled solutions with organizations in a variety of verticals. Prior to joining Mosaic Data Science, Michael led the development of analytics solutions that enabled new "smart office" Internet-of-Things-based products and services at Steelcase and researched air-traffic management at the NASA Ames Research Center. He received his B.S.E. degree with majors in electrical and computer engineering and economics from Calvin College in 2004 and his M.S. degree in electrical and computer engineering from the University of Illinois at Urbana-Champaign in 2007. In 2015, Michael received his PhD in operations research from the Department of Management Science & Engineering at Stanford University.