I work on inverse optimization problems – identifying if a decision process is a solution to an optimization problem, and if so, estimating the parameters of the optimization problem. This problem is well-known in two domains: inverse reinforcement learning in machine learning and revealed preference in economics. For my research, I draw on results from both of these areas. My primary contributions are three-fold:
(1) I develop novel inverse optimization algorithms for Bayesian decision makers that do not require any knowledge of the decision maker’s model dynamics by exploiting problem structure.
(2) I conduct fundamental research in economics theory and develop algorithms that provably preserve decision privacy from adversarial eavesdroppers.
(3) On the applied side of research, I work on interpretable machine learning and develop economics-based interpretable models that explain the behavior of neural networks. I also analyze social multimedia datasets (like YouTube) and predict user engagement. Finally, I work with defense industries to develop system-level algorithms for cognitive radars in adversarial settings.
(More on the way)