Score Matching with Missing Data

Josh Givens, Song Liu, Henry Reeve

International Conference on Machine Learning 2025 · Oral

This talk, presented by Song Liu and spearheaded by his PhD student Josh Givens, introduces a novel framework for applying **score matching** techniques to datasets afflicted by missing values. Titled "Score Matching with Missing Data," the research addresses a critical gap in the application of powerful generative models and statistical inference methods to real-world, incomplete data. Score matching is a foundational component of many modern machine learning paradigms, including **diffusion models** and **energy-based models**, which are increasingly vital for tasks like high-fidelity data generation, image synthesis, and robust inference.

AI review

A competent and honestly scoped extension of score matching to the MCAR missing data setting. The Marginal Score Matching criterion is a natural formulation, the IW branch delivers a genuine finite-sample consistency result, and the MVSM branch offers a practically motivated variational approximation. The theoretical gap between the two branches — one proven, one not — is the paper's honest weak point, and the empirical validation is thin by current standards. This is solid, publishable work that will be useful to a specific community, but it is not the kind of result that reframes the…