Game-Theoretic Algorithms for Conditional Moment Matching
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. and Dai et al. as special cases of our general framework before detailing various extensions and how to efficiently solve the game defined by CMR.
Code (0)
등록된 구현이 없습니다.
Tasks
EconometricsregressionSimilar Papers 제목 키워드 기반
Of Moments and Matching: A Game-Theoretic Framework for Closing the Imitation Gap
We provide a unifying view of a large family of previous imitation learning algorithms through the lens of moment matching. At its core, our classification scheme is based on whether the learner attempts to match (1) rew…
Imitation LearningRapid Learning in Constrained Minimax Games with Negative Momentum
In this paper, we delve into the utilization of the negative momentum technique in constrained minimax games. From an intuitive mechanical standpoint, we introduce a novel framework for momentum buffer updating, which ex…
FormContinuous-Time Analysis of Heavy Ball Momentum in Min-Max Games
Since Polyak's pioneering work, heavy ball (HB) momentum has been widely studied in minimization. However, its role in min-max games remains largely unexplored. As a key component of practical min-max algorithms like Ada…
The Variational Method of Moments
The conditional moment problem is a powerful formulation for describing structural causal parameters in terms of observables, a prominent example being instrumental variable regression. A standard approach reduces the pr…
validDensity Approximation of Affine Jump Diffusions via Closed-Form Moment Matching
We develop a recursive approach for deriving closed-form solutions to both conditional and unconditional moments of affine jump diffusions with state-independent jump intensities. Using these moment solutions, we constru…
Computational EfficiencyForm