Low-rank Distributional Matrix Completion

In three linesNovel matrix completion framework where each entry is a probability distribution rather than a scalar. Uses kernel mean embeddings and introduces Tucker rank notion for distribution-valued matrices. Proposes estimator with non-asymptotic error bounds and experimental validation on synthetic and real-world data.

Summary generated by Claude — human-verified