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

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

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

Computer Science > Machine Learning

Title: Factor Group-Sparse Regularization for Efficient Low-Rank Matrix Recovery

Abstract: This paper develops a new class of nonconvex regularizers for low-rank matrix recovery. Many regularizers are motivated as convex relaxations of the matrix rank function. Our new factor group-sparse regularizers are motivated as a relaxation of the number of nonzero columns in a factorization of the matrix. These nonconvex regularizers are sharper than the nuclear norm; indeed, we show they are related to Schatten-$p$ norms with arbitrarily small $0 < p \leq 1$. Moreover, these factor group-sparse regularizers can be written in a factored form that enables efficient and effective nonconvex optimization; notably, the method does not use singular value decomposition. We provide generalization error bounds for low-rank matrix completion which show improved upper bounds for Schatten-$p$ norm reglarization as $p$ decreases. Compared to the max norm and the factored formulation of the nuclear norm, factor group-sparse regularizers are more efficient, accurate, and robust to the initial guess of rank. Experiments show promising performance of factor group-sparse regularization for low-rank matrix completion and robust principal component analysis.
Comments: Accepted by NeurIPS 2019. The supplementary material is at this https URL
Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML)
Cite as: arXiv:1911.05774 [cs.LG]
  (or arXiv:1911.05774v2 [cs.LG] for this version)

Submission history

From: Jicong Fan [view email]
[v1] Wed, 13 Nov 2019 19:30:35 GMT (90kb)
[v2] Mon, 18 Nov 2019 16:30:50 GMT (90kb)

Link back to: arXiv, form interface, contact.