Doubly Robust Deep Kernel Tests for Structured Outcomes
Abstract
We propose a test of whether two interventions induce the same outcome distribution when outcomes are structured and high-dimensional, such as images, text and graphs. We represent the interventional distributions as mean embeddings on learned deep features, such that the treatment effect is a doubly robust representation of the difference in mean embeddings. Our test, DR-KTE-Deep, optimises the learned features so as to maximise an estimated power criterion on this treatment effect representation. To motivate our learning objective, we show that the power criterion characterises the asymptotic signal-to-noise ratio of the test under fixed alternatives, and we derive a uniform finite-sample bound for its regularised estimate. We show that, with the learned kernel fixed, a held-out test is asymptotically valid under the null given suitable nuisance-estimation rates. Empirically, DR-KTE-Deep achieves substantially higher power than fixed and bandwidth-optimised kernel baselines across image, text and graph outcomes, while maintaining well-calibrated Type-I error.
Related
- Flexible Doubly Robust Proximal Causal Inference for Censored Outcomes
- Win-Ratio Regression for Prioritized Composite Outcomes in Observational Studies: Doubly Robust and Efficient Estimation with Future-Score Correction
- Doubly Robust Estimation and Inference for Longitudinal Win Probabilities with Ordinal Outcomes in Observational Studies
- Anytime-Valid Inference in Linear Models with Clustered Outcomes: Exact Multivariate Tests and Cluster-Robust Confidence Sequences
- A Doubly Robust Approach to Evaluating the Impact of Hospital Length of Stay on Post‐Fracture Recovery in Older Adults With Cognitive Impairment
BibTeX
@misc{leman2026doubly,
title = {{Doubly Robust Deep Kernel Tests for Structured Outcomes}},
author = {Robin Leman and Bariscan Bozkurt and Sofiane Tanji and Arthur Gretton and Houssam Zenati},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.37227},
eprint = {2610.37227},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.37227},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Doubly Robust Deep Kernel Tests for Structured Outcomes
AU - Robin Leman
AU - Bariscan Bozkurt
AU - Sofiane Tanji
AU - Arthur Gretton
AU - Houssam Zenati
PY - 2026
DA - 2026/10/27
PB - arXiv
DO - 10.48550/arxiv.2610.37227
UR - https://arxiv.org/abs/2610.37227
AB - We propose a test of whether two interventions induce the same outcome distribution when outcomes are structured and high-dimensional, such as images, text and graphs. We represent the interventional distributions as mean embeddings on learned deep features, such that the treatment effect is a doubly robust representation of the difference in mean embeddings. Our test, DR-KTE-Deep, optimises the learned features so as to maximise an estimated power criterion on this treatment effect representation. To motivate our learning objective, we show that the power criterion characterises the asymptotic signal-to-noise ratio of the test under fixed alternatives, and we derive a uniform finite-sample bound for its regularised estimate. We show that, with the learned kernel fixed, a held-out test is asymptotically valid under the null given suitable nuisance-estimation rates. Empirically, DR-KTE-Deep achieves substantially higher power than fixed and bandwidth-optimised kernel baselines across image, text and graph outcomes, while maintaining well-calibrated Type-I error.
ER -