I conduct research in the areas of Industrial Organization, Inverse Reinforcement Learning, and Reinforcement Learning.

IRL/RL Research

Inverse Reinforcement Learning with Conditional Choice Probabilities

Mohit Sharma, Kris M. Kitani, Joachim Groeger

We make an important connection to existing results in econometrics to describe an alternative formulation of inverse reinforcement learning (IRL). In particular, we describe an algorithm using Conditional Choice Probabilities (CCP), which are maximum likelihood estimates of the policy estimated from expert demonstrations, to solve the IRL problem. Using the language of structural econometrics, we re-frame the optimal decision problem and introduce an alternative representation of value functions due to (Hotz and Miller 1993). In addition to presenting the theoretical connections that bridge the IRL literature between Economics and Robotics, the use of CCPs also has the practical benefit of reducing the computational cost of solving the IRL problem. Specifically, under the CCP representation, we show how one can avoid repeated calls to the dynamic programming subroutine typically used in IRL. We show via extensive experimentation on standard IRL benchmarks that CCP-IRL is able to outperform MaxEnt-IRL, with as much as a 5x speedup and without compromising on the quality of the recovered reward function.

Applied Micro Research

A Study of Participation in Dynamic Auctions

Published in Vol. 55, No. 4, November 2014,  International Economic Review

This paper studies repeated entry and bidding decisions in construction procurement auctions.

Draft Version

Bidding frictions in Ascending Auctions with Aaron Barkley and Robert A. Miller (supersedes "Ascending Auctions with Costly monitoring")

This paper develops a strategy for identifying and estimating the distribution of valuations in ascending auctions where bidders have an unknown number of bidding opportunities. While the model may have multiple equilibria and equilibrium strategies may be non-monotone in bidder valuations, we use a restriction to undominated strategies to identify the valuation distribution. The model is applied to a monthly financial market in which local banks compete for deposit securities issued by the state of Texas. Several features of the data suggest standard empirical methods for modelling ascending auctions may not be suitable for this market, including frequent jump bidding and winning bids well above the highest losing bid. We find that frictions are costly both for revenue and allocative efficiency. 


The informational content of the limit order book: an empirical study of prediction markets

In this paper I empirically investigate prediction markets for binary options (Arrow-Debreu securities) for political events.