arXiv:2607.07767v1 Announce Type: cross 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 semi definite (PSD) kernel densities we obtain a convex empirical risk problem with closed form marginals, solvable by a Newton inte
The increasing complexity and scale of AI/ML models necessitate more robust and theoretically sound approaches to data handling, especially concerning prevalent missing data scenarios.
Improved imputation methods reduce bias and strengthen the reliability of machine learning models across various applications, making AI systems more trustworthy and effective.
Machine learning pipelines can now incorporate a more theoretically grounded and robust method for handling missing data, moving beyond heuristic or restrictive parametric assumptions.
- · Machine Learning Researchers
- · Data Scientists
- · Industries relying on large datasets with missing values
- · AI development in general
- · Methods relying on simple imputation heuristics
- · AI applications with high data sensitivity and poor data quality
More accurate and reliable AI models due to better handling of incomplete datasets.
Reduced need for expensive and time-consuming manual data cleaning and preparation in certain AI applications.
Accelerated development and adoption of AI in domains where data completeness is a major challenge.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG