ADVERSARIALLY BALANCED REPRESENTATION FOR CONTINUOUS TREATMENT EFFECT ESTIMATION

Amirreza Kazemi, Martin Ester

Abstract:

Estimating the individual treatment effect (ITE) requires covariate balance among different treatment groups, and machine learning models have shown great promise in learning a balanced representation of covariates.
In contrast with binary treatments for which learning such a representation has been widely studied, the more practical yet complicated continuous treatment setting has remained relatively under-explored. Adopting an information-theoretic approach, we introduce a novel mutual information (MI)-based objective for continuous treatment effect estimation.
Leveraging variational approximation to optimize MI terms in our objective, we propose a method called Adversarial CounterFactual Regression (ACFR). ACFR aligns the representation of covariates through an adversarial game and predicts the potential outcomes using a contribution-constraining hypothesis network. Comparison of ACFR against state-of-the-art methods on semi-synthetic datasets demonstrates its superiority in individual-level metrics.

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

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/adversarially-balanced-representation-for/code)