File Download
There are no files associated with this item.
Supplementary
-
Citations:
- Scopus: 0
- Appears in Collections:
Article: Non-convex matrix completion and related problems via strong duality
Title | Non-convex matrix completion and related problems via strong duality |
---|---|
Authors | |
Keywords | Matrix completion Matrix factorization Non-convex optimization Robust principal component analysis Sample complexity Strong duality |
Issue Date | 2019 |
Citation | Journal of Machine Learning Research, 2019, v. 20 How to Cite? |
Abstract | This work studies the strong duality of non-convex matrix factorization problems: we show that under certain dual conditions, these problems and the dual have the same optimum. This has been well understood for convex optimization, but little was known for non-convex problems. We propose a novel analytical framework and prove that under certain dual conditions, the optimal solution of the matrix factorization program is the same as that of its bi-dual and thus the global optimality of the non-convex program can be achieved by solving its bi-dual which is convex. These dual conditions are satisfied by a wide class of matrix factorization problems, although matrix factorization is hard to solve in full generality. This analytical framework may be of independent interest to non-convex optimization more broadly. We apply our framework to two prototypical matrix factorization problems: Matrix completion and robust Principal Component Analysis. These are examples of efficiently recovering a hidden matrix given limited reliable observations. Our framework shows that exact recoverability and strong duality hold with nearly-optimal sample complexity for the two problems. |
Persistent Identifier | http://hdl.handle.net/10722/341255 |
ISSN | 2023 Impact Factor: 4.3 2023 SCImago Journal Rankings: 2.796 |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Balcan, Maria Florina | - |
dc.contributor.author | Liang, Yingyu | - |
dc.contributor.author | Song, Zhao | - |
dc.contributor.author | Woodruff, David P. | - |
dc.contributor.author | Zhang, Hongyang | - |
dc.date.accessioned | 2024-03-13T08:41:23Z | - |
dc.date.available | 2024-03-13T08:41:23Z | - |
dc.date.issued | 2019 | - |
dc.identifier.citation | Journal of Machine Learning Research, 2019, v. 20 | - |
dc.identifier.issn | 1532-4435 | - |
dc.identifier.uri | http://hdl.handle.net/10722/341255 | - |
dc.description.abstract | This work studies the strong duality of non-convex matrix factorization problems: we show that under certain dual conditions, these problems and the dual have the same optimum. This has been well understood for convex optimization, but little was known for non-convex problems. We propose a novel analytical framework and prove that under certain dual conditions, the optimal solution of the matrix factorization program is the same as that of its bi-dual and thus the global optimality of the non-convex program can be achieved by solving its bi-dual which is convex. These dual conditions are satisfied by a wide class of matrix factorization problems, although matrix factorization is hard to solve in full generality. This analytical framework may be of independent interest to non-convex optimization more broadly. We apply our framework to two prototypical matrix factorization problems: Matrix completion and robust Principal Component Analysis. These are examples of efficiently recovering a hidden matrix given limited reliable observations. Our framework shows that exact recoverability and strong duality hold with nearly-optimal sample complexity for the two problems. | - |
dc.language | eng | - |
dc.relation.ispartof | Journal of Machine Learning Research | - |
dc.subject | Matrix completion | - |
dc.subject | Matrix factorization | - |
dc.subject | Non-convex optimization | - |
dc.subject | Robust principal component analysis | - |
dc.subject | Sample complexity | - |
dc.subject | Strong duality | - |
dc.title | Non-convex matrix completion and related problems via strong duality | - |
dc.type | Article | - |
dc.description.nature | link_to_subscribed_fulltext | - |
dc.identifier.scopus | eid_2-s2.0-85072624374 | - |
dc.identifier.volume | 20 | - |
dc.identifier.eissn | 1533-7928 | - |