Sinkhorn Discrepancy for Counterfactual Generalization

Hao Wang, Quanyu Dai, Jiajun Fan, Weiming Liu, Zhichao Chen, Tianqiao Liu, Yichao Wang, Zhenhua Dong, Ruiming Tang

Keywords: causal inference, treatment selection bias

Abstract:

Estimating individual treatment effects from observational data is very challenging due to the existence of treatment selection bias.
Most existing representation-based methods mitigate this issue by aligning distributions of different treatment groups in the representation space. However, they still suffer from two critical problems: (1) Mini-batch Sampling Effects (MSE), where the alignment easily fails due to the outcome imbalance or outliers in the batch; (2) Unobserved Confounder Effects (UCE), where the unobserved confounders damage the correct alignment. To tackle these problems, we propose a principled approach named Entire Space CounterFactual Regression (ESCFR) based on a generalized sinkhorn discrepancy for distribution alignment within the stochastic optimal transport framework. Based on the framework, we propose a relaxed mass preserving regularizer to address the MSE issue and design a proximal factual outcome regularizer to handle the UCE issue. Extensive experiments demonstrate that our proposed ESCFR can successfully tackle the treatment selection bias and achieve significantly better performance than state-of-the-art methods.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning