We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: Beyond the Signs: Nonparametric Tensor Completion via Sign Series

Abstract: We consider the problem of tensor estimation from noisy observations with possibly missing entries. A nonparametric approach to tensor completion is developed based on a new model which we coin as sign representable tensors. The model represents the signal tensor of interest using a series of structured sign tensors. Unlike earlier methods, the sign series representation effectively addresses both low- and high-rank signals, while encompassing many existing tensor models -- including CP models, Tucker models, single index models, several hypergraphon models -- as special cases. We show that the sign tensor series is theoretically characterized, and computationally estimable, via classification tasks with carefully-specified weights. Excess risk bounds, estimation error rates, and sample complexities are established. We demonstrate the outperformance of our approach over previous methods on two datasets, one on human brain connectivity networks and the other on topic data mining.
Comments: 28 pages, 6 figures
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Statistics Theory (math.ST); Methodology (stat.ME)
Journal reference: 35th Annual Conference on Neural Information Processing Systems (NeurIPS 2021)
Cite as: arXiv:2102.00384 [stat.ML]
  (or arXiv:2102.00384v1 [stat.ML] for this version)

Submission history

From: Chanwoo Lee [view email]
[v1] Sun, 31 Jan 2021 05:27:01 GMT (1196kb,D)

Link back to: arXiv, form interface, contact.