Learning Disentangled Representations for Counterfactual Regression via Mutual Information Minimization

TL;DR

This paper introduces MIM-DRCFR, a novel method that employs mutual information minimization within a multi-task framework to learn truly disentangled, independent latent factors for more accurate individual treatment effect estimation.

Contribution

It proposes a new approach combining MI minimization with multi-task learning to improve the independence of latent factors in disentangled representation learning for causal inference.

Findings

Outperforms state-of-the-art methods on benchmarks

Effective in real-world industrial datasets

Ensures independence of latent factors

Abstract

Learning individual-level treatment effect is a fundamental problem in causal inference and has received increasing attention in many areas, especially in the user growth area which concerns many internet companies. Recently, disentangled representation learning methods that decompose covariates into three latent factors, including instrumental, confounding and adjustment factors, have witnessed great success in treatment effect estimation. However, it remains an open problem how to learn the underlying disentangled factors precisely. Specifically, previous methods fail to obtain independent disentangled factors, which is a necessary condition for identifying treatment effect. In this paper, we propose Disentangled Representations for Counterfactual Regression via Mutual Information Minimization (MIM-DRCFR), which uses a multi-task learning framework to share information when learning the latent factors and incorporates MI minimization learning criteria to ensure the independence of these factors. Extensive experiments including public benchmarks and real-world industrial user growth datasets demonstrate that our method performs much better than state-of-the-art methods.