Distributionally Faithful Imputation via Positive Semi-Definite Kernel Density Estimation
Abstract
Missing values undermine statistical inference and machine learning pipelines, yet most imputation methods rely on heuristics or restrictive parametric assumptions that ignore the joint data distribution. We recast imputation under missing completely at random (MCAR) as density estimation from masked observations: estimate a distribution whose observed marginals exactly match those in the data. Leveraging positive semidefinite (PSD) kernel densities we obtain a convex empirical risk problem with closed form marginals, solvable by a Newton interior point method. We further introduce a non convex extension that supports adaptive hyperparameter selection through alternating minimization. The resulting PSD-Impute model yields both single and multiple imputations from the same fitted density, enjoys statistical consistency, beating the curse of dimensionality for very regular probabilities. Experiments indicate competitive distributional accuracy compared with popular imputation baselines, suggesting strong practical promise.
Related
- A positive-definite Fourier transform, with applications to fractional Brownian motions and Bayesian probabilistic numerics
- Geometric Feature Learning for Functional Data Valued on the Symmetric Positive Definite Manifold
- Geodesics and intrinsic statistics of isolated connections in Hermitian positive-definite matrices
- Structured Positive-Definite Optimal Control Synthesis of Closed-Loop Recommendation Systems over Social Networks
- Conservative neural posterior estimation via distributionally robust training
BibTeX
@misc{basteri2026distributionally,
title = {{Distributionally Faithful Imputation via Positive Semi-Definite Kernel Density Estimation}},
author = {Andrea Basteri and Carlo Ciliberto and Alessandro Rudi},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2607.07767},
eprint = {2607.07767},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2607.07767},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Distributionally Faithful Imputation via Positive Semi-Definite Kernel Density Estimation
AU - Andrea Basteri
AU - Carlo Ciliberto
AU - Alessandro Rudi
PY - 2026
DA - 2026/10/27
PB - arXiv
DO - 10.48550/arxiv.2607.07767
UR - https://arxiv.org/abs/2607.07767
AB - Missing values undermine statistical inference and machine learning pipelines, yet most imputation methods rely on heuristics or restrictive parametric assumptions that ignore the joint data distribution. We recast imputation under missing completely at random (MCAR) as density estimation from masked observations: estimate a distribution whose observed marginals exactly match those in the data. Leveraging positive semidefinite (PSD) kernel densities we obtain a convex empirical risk problem with closed form marginals, solvable by a Newton interior point method. We further introduce a non convex extension that supports adaptive hyperparameter selection through alternating minimization. The resulting PSD-Impute model yields both single and multiple imputations from the same fitted density, enjoys statistical consistency, beating the curse of dimensionality for very regular probabilities. Experiments indicate competitive distributional accuracy compared with popular imputation baselines, suggesting strong practical promise.
ER -