Join us in welcoming Archis Ghate, ISE department head at the University of Minnesota, as he discusses Inverse Markov Decision Processes. Alums and friends of the program are always welcome.

Did You Miss It?

That’s okay. You can Watch it Here.
Access Password: ^CRm8^B@

Title and Abstract

Inverse Markov Decision Processes

Inverse optimization seeks parameter values that make the chosen decision variables optimal. In contrast, forward optimization computes optimal decision variables from given parameters. This talk will focus on inverse optimization in Markov decision processes (MDPs). Specifically, we will study how to impute rewards and transition probabilities that make a policy optimal.

To start, we will review prior work on imputing rewards. After that, we will formulate the problem of imputing transition probabilities. While both issues rely on mathematical programming forms of Bellman’s equations, they differ in difficulty. In particular, imputing rewards is easier because the problem is linear. However, imputing transition probabilities is harder since the issue is bilinear. Therefore, we will present exact and approximate methods for solving versions of this bilinear challenge.

Next, we will show how to use uniformization to extend these methods to continuous-time MDPs. Finally, we will also describe an extension to semi-Markov decision processes.

Biographies

Archis Ghate is a professor and the department head of industrial and systems engineering at the University of Minnesota. Previously, he worked as a professor of industrial engineering and held the Fluor Endowed Chair at Clemson University.

Before joining Clemson, Ghate was an industrial and systems engineering professor at the University of Washington in Seattle. During that time, he worked as associate chair for six years. In addition, he held a College of Engineering endowed professorship for five years.

He began his career at the University of Washington as an assistant professor. He earned a Ph.D. in industrial and operations engineering from the University of Michigan in 2006. He also received an M.S. in management science and engineering from Stanford in 2003. Moreover, he completed his undergraduate studies at the Indian Institute of Technology, Bombay, in 2001.

Ghate has received many honors. For example, he earned the NSF CAREER Award. He also won an excellence in teaching operations research award and a best paper award from IISE. Furthermore, he has received several teaching accolades from the University of Washington. His students have also excelled by winning the Dantzig Dissertation Award and the Bonder Scholarship in healthcare operations research from INFORMS.

In addition, Ghate has served on the editorial boards of several journals. He also held leadership roles, including general chair of the INFORMS 2019 Annual Meeting and program co-chair of the 2021 IISE Annual Conference.