• Corpus ID: 235367652

Deep Proxy Causal Learning and its Application to Confounded Bandit Policy Evaluation

  title={Deep Proxy Causal Learning and its Application to Confounded Bandit Policy Evaluation},
  author={Liyuan Xu and Heishiro Kanagawa and Arthur Gretton},
  booktitle={Neural Information Processing Systems},
Proxy causal learning (PCL) is a method for estimating the causal effect of treatments on outcomes in the presence of unobserved confounding, using proxies (structured side information) for the confounder. This is achieved via two-stage regression: in the first stage, we model relations among the treatment and proxies; in the second stage, we use this model to learn the effect of treatment on the outcome, given the context provided by the proxies. PCL guarantees recovery of the true causal… 

Figures and Tables from this paper

Deep Learning Methods for Proximal Inference via Maximum Moment Restriction

This work introduces a flexible and scalable method based on a deep neural network to estimate causal effects in the presence of unmeasured confounding using proximal inference and provides theoretical consistency guarantees for this method.

A Minimax Learning Approach to Off-Policy Evaluation in Confounded Partially Observable Markov Decision Processes

We consider off-policy evaluation (OPE) in Partially Observable Markov Decision Processes (POMDPs), where the evaluation policy depends only on observable variables and the behavior policy depends on

A Minimax Learning Approach to Off-Policy Evaluation in Partially Observable Markov Decision Processes

This work proposes novel identification methods for OPE in POMDPs with latent confounders, by introducing bridge functions that link the target policy’s value and the observed data distribution, and proposes minimax estimation methods for learning these bridge functions.

Dual Instrumental Method for Confounded Kernelized Bandits

To deal with the challenges brought by the confounder, this work applies the dual instrumental variable regression, which can correctly identify the true reward function and proves the convergence rate of this method is near-optimal in two types of widely used reproducing kernel Hilbert spaces.

Proximal Reinforcement Learning: Efficient Off-Policy Evaluation in Partially Observed Markov Decision Processes

This work considers off-policy evaluation in a partially observed MDP (POMDP) by considering estimating the value of a given target policy in a POMDP given trajectories with only partial state observations generated by a different and unknown policy that may depend on the unobserved state.

Nonlinear Causal Discovery via Kernel Anchor Regression

This work tackles the nonlinear setting by proposing kernel anchor regression (KAR), and beyond the natural formula-tion using a classic two-stage least square estimator, also study an improved variant that involves nonparametric regression in three separate stages.

Long-term Causal Inference Under Persistent Confounding via Data Combination

We study the identification and estimation of long-term treatment effects when both experimental and observational data are available. Since the long-term outcome is observed only after a long delay,

Estimating individual treatment effects under unobserved confounding using binary instruments

This paper proposes a novel, multiply robust machine learning framework, called MRIV, for estimating ITEs using binary IVs and demonstrates empirically that the MRIV-Net achieves state-of-the-art performance.

A Neural Mean Embedding Approach for Back-door and Front-door Adjustment

The proposed method is shown to converge to the true causal parameter, and outperforms the recent state-of-the-art methods on challenging causal benchmarks, including settings involving high-dimensional image data.

Statistical Estimation of Confounded Linear MDPs: An Instrumental Variable Approach

In an Markov decision process (MDP), unobservable confounders may exist and have impacts on the data generating process, so that the classic off-policy evaluation (OPE) estimators may fail to identify



Proximal Causal Learning with Kernels: Two-Stage Estimation and Moment Restriction

This work proposes two kernel-based methods for nonlinear causal effect estimation in the presence of unobserved confounding, but where proxies for the latent confounder(s) are observed: a two-stage regression approach, and a maximum moment restriction approach.

Causal Effect Inference with Deep Latent-Variable Models

This work builds on recent advances in latent variable modeling to simultaneously estimate the unknown latent space summarizing the confounders and the causal effect and shows its method is significantly more robust than existing methods, and matches the state-of-the-art on previous benchmarks focused on individual treatment effects.

Learning Deep Features in Instrumental Variable Regression

Deep feature instrumental variable regression (DFIV) is proposed, to address the case where relations between instruments, treatments, and outcomes may be nonlinear, and outperforms recent state-of-the-art methods on challenging IV benchmarks, including settings involving high dimensional image data.

Representation Learning for Treatment Effect Estimation from Observational Data

A local similarity preserved individual treatment effect (SITE) estimation method based on deep representation learning that preserves local similarity and balances data distributions simultaneously, by focusing on several hard samples in each mini-batch.

An Introduction to Proximal Causal Learning

A formal potential outcome framework for proximal causal learning is introduced, which while explicitly acknowledging covariate measurements as imperfect proxies of confounding mechanisms, offers an opportunity to learn about causal effects in settings where exchangeability on the basis of measured covariates fails.

Provably Efficient Neural Estimation of Structural Equation Model: An Adversarial Approach

For the first time, this work provides a tractable estimation procedure for SEMs based on NNs with provable convergence and without the need for sample splitting.

Deep Generalized Method of Moments for Instrumental Variable Analysis

This paper proposes the DeepGMM algorithm, a new variational reformulation of GMM with optimal inverse-covariance weighting that allows us to efficiently control very many moment conditions and develops practical techniques for optimization and model selection that make it particularly successful in practice.

Estimation of Individual Treatment Effect in Latent Confounder Models via Adversarial Learning

A novel adversarial learning framework is developed to make unbiased estimates of the ITE using noisy proxies to handle confounders in observational data.

Proxy Controls and Panel Data.

A flexible approach to the identification and estimation of causal objects in nonparametric, non-separable models with confounding with the use of `proxy controls': covariates that do not satisfy a standard `unconfoundedness' assumption but which are informative proxies for variables that do.

A Proxy Variable View of Shared Confounding

This paper focuses on the setting where there are many treatments with shared confounding, and it studies under what conditions is causal identification possible, and shows that one can circumvent the need for such solutions by directly modeling the data.