Get Started
Research questionWhen are Double Machine Learning estimators inadmissible under structure-agnostic models?Structure-agnostic models permit observed-data laws near fixed black-box machine-learning estimates without specifying conventional structural assumptions. Although an estimator can be minimax, another estimator may achieve uniformly better asymptotic risk for particular functionals.
Machine Learning
Research Paper
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.On the Asymptotic Inadmissibility of Double Machine Learning Estimators Under Structure-Agnostic ModelsThe analysis covers the quadratic functional in the Gaussian sequence model, the quadratic density integral, and expected conditional covariance. It finds asymptotic inadmissibility of DML for the first two through empirical higher-order influence-function estimators, while neither estimator asymptotically dominates the other for expected conditional covariance. The structure-agnostic neighborhoods encode convergence rates of the black-box estimates.research paper · Sep 4, 2026
Related questions
Can language-based models replace specialized architectures for structured data without sacrificing structural representation and computation?How can smooth density estimation preserve accuracy under relaxed local differential privacy?Can one squared-loss estimator achieve both minimax and universal exponential rates for finite versus countably infinite hypothesis classes?How can LLMs distinguish ambiguous inputs from gaps in their knowledge when estimating uncertainty?
Home
Topics
Search
Library