Toward Scalable and Valid Conditional Independence Testing with Spectral Representations
Not on the map. Why?
Abstract (arXiv)
Conditional independence (CI) is central to causal inference, feature selection, and graphical modeling, yet it is untestable in many settings without additional assumptions. Existing CI tests often rely on restrictive structural conditions, limiting their validity. Kernel methods using partial covariance operators offer a more principled approach but suffer from limited adaptivity and scalability. In this work, we explore whether representation learning can help address these limitations. Specifically, we focus on representations derived from the singular value decomposition of partial covariance operators and use them to construct a simple test statistic. We also introduce a bi-level contrastive algorithm to learn these representations. Our theory links representation learning error to test performance and establishes asymptotic validity and power guarantees. Experiments on real and synthetic data suggest that this approach offers a principled and statistically grounded path toward scalable CI testing, bridging kernel-based theory with modern representation learning.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 29 of its 75 references that are in the map. First posted on arXiv 22 Dec 2025.
Builds on
29 of 75 references.
References · 29
- 2016XGBoost: A Scalable Tree Boosting Systemcited by 5,166
- 2016Kernel Mean Embedding of Distributions: A Review and Beyondcited by 364
- 2021Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive Losscited by 313
- 2016Panning for Gold: Model-X Knockoffs for High-dimensional Controlled Variable Selectioncited by 121
- 2017Approximate Kernel-based Conditional Independence Tests for Fast Non-Parametric Causal Discoverycited by 76
- 2017Conditional independence testing based on a nearest-neighbor estimator of conditional mutual informationcited by 70
- 2017Model-Powered Conditional Independence Testcited by 43
- 2019Conditional Independence Testing using Generative Adversarial Networkscited by 36
- 2021Local permutation tests for conditional independencecited by 23
- 2017Nonparametric independence testing via mutual informationcited by 22
- 2020Double Generative Adversarial Networks for Conditional Independence Testingcited by 22
- 2020Kernel Partial Correlation Coefficient -- a Measure of Conditional Dependencecited by 21
17 more
- 2022Efficient Conditionally Invariant Representation Learningcited by 19
- 2024Practical Kernel Tests of Conditional Independencecited by 15
- 2024Neural Conditional Probability for Uncertainty Quantificationcited by 14
- 2021An Asymptotic Test for Conditional Independence using Analytic Kernel Embeddingscited by 12
- 2022Spectral Representation Learning for Conditional Moment Modelscited by 12
- 2024Spectral Representation for Causal Estimation with Hidden Confounderscited by 12
- 2023Domain-specific optimization and diverse evaluation of self-supervised models for histopathologycited by 11
- 2024Conditional Diffusion Models Based Conditional Independence Testingcited by 9
- 2024Doubly Robust Conditional Independence Testing with Generative Neural Networkscited by 7
- 2025Self-Supervised Evolution Operator Learning for High-Dimensional Dynamical Systemscited by 7
- 2025On the Hardness of Conditional Independence Testing In Practicecited by 7
- 2025Position: Causal Machine Learning Requires Rigorous Synthetic Experiments for Broader Adoptioncited by 6
- 2020On the power of conditional independence testing under model-Xcited by 4
- 2024Primal-Dual Spectral Representation for Off-policy Evaluationcited by 4
- 2025Demystifying Spectral Feature Learning for Instrumental Variable Regressioncited by 4
- 2025Representation Learning for Equivariant Inference with Guaranteescited by 3
- 2025Outcome-Aware Spectral Feature Learning for Instrumental Variable Regressioncited by 2
Built on by
Citing papers per year: 2026: 1.
1 paper.