Game-Theoretic Algorithms for Conditional Moment Matching - Robotics Institute Carnegie Mellon University

Game-Theoretic Algorithms for Conditional Moment Matching

Tech. Report, CMU-RI-TR-22-53, Robotics Institute, Carnegie Mellon University, August, 2022

Abstract

A variety of problems in econometrics and machine learning, including instrumental variable regression and Bellman residual minimization, can be formulated as satisfying a set of conditional moment restrictions (CMR). We derive a general, game-theoretic strategy for satisfying CMR that scales to nonlinear problems, is amenable to gradient-based optimization, and is able to account for finite sample uncertainty. We recover the approaches of (Dikkala et al., 2020) and (Dai et al., 2018) as special cases of our general framework before detailing various extensions and how to efficiently solve the game defined by CMR.

BibTeX

@techreport{Swamy-2022-133115,
author = {Gokul Swamy and Sanjiban Choudhury and J. Andrew Bagnell and Zhiwei Steven Wu},
title = {Game-Theoretic Algorithms for Conditional Moment Matching},
year = {2022},
month = {August},
institute = {Carnegie Mellon University},
address = {Pittsburgh, PA},
number = {CMU-RI-TR-22-53},
keywords = {conditional moment restriction instrumental variable regression reinforcement learning},
}