Theoretical Analysis of Measure Consistency Regularization for Partially Observed Data

Yinsong Wang, and Shahin Shahrampour · 2026

Read Theoretical Analysis of Measure Consistency Regularization for Partially Observed Data by Yinsong Wang, and Shahin Shahrampour (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-01437.txt.

Description

Yinsong Wang and Shahin Shahrampour study Measure Consistency Regularization, forcing agreement between imputed and observed data to aid generalization. Such regularization has succeeded for inpainting, imputation, and semi-supervised learning, yet theory lagged. Through neural network distance, they pin down when it gives a better error bound. Under ideal interpolation and compatibility, its bound beats vanilla supervised training, strictly beating it when mixed samples win. In non-ideal settings, residuals can erase that edge. A diagnostic pairing a duality-gap check with a drift estimator judges whether such training helps.

Catalog record

CREATORYinsong Wang, and Shahin Shahrampour
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE96128
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.01437

Browse the full library · About the Tower of Babel