FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4224256927> ?p ?o ?g. }

• W4224256927 abstract "Low-rank matrix estimation under heavy-tailed noise is challenging, both computationally and statistically. Convex approaches have been proven statistically optimal but suffer from high computational costs, especially since robust loss functions are usually non-smooth. More recently, computationally fast non-convex approaches via sub-gradient descent are proposed, which, unfortunately, fail to deliver a statistically consistent estimator even under sub-Gaussian noise. In this paper, we introduce a novel Riemannian sub-gradient (RsGrad) algorithm which is not only computationally efficient with linear convergence but also is statistically optimal, be the noise Gaussian or heavy-tailed. Convergence theory is established for a general framework and specific applications to absolute loss, Huber loss, and quantile loss are investigated. Compared with existing non-convex methods, ours reveals a surprising phenomenon of dual-phase convergence. In phase one, RsGrad behaves as in a typical non-smooth optimization that requires gradually decaying stepsizes. However, phase one only delivers a statistically sub-optimal estimator which is already observed in the existing literature. Interestingly, during phase two, RsGrad converges linearly as if minimizing a smooth and strongly convex objective function and thus a constant stepsize suffices. Underlying the phase-two convergence is the smoothing effect of random noise to the non-smooth robust losses in an area close but not too close to the truth. Lastly, RsGrad is applicable for low-rank tensor estimation under heavy-tailed noise where a statistically optimal rate is attainable with the same phenomenon of dual-phase convergence, and a novel shrinkage-based second-order moment method is guaranteed to deliver a warm initialization. Numerical simulations confirm our theoretical discovery and showcase the superiority of RsGrad over prior methods." @default.
• W4224256927 created "2022-04-26" @default.
• W4224256927 creator A5034178721 @default.
• W4224256927 creator A5036610839 @default.
• W4224256927 creator A5043467180 @default.
• W4224256927 creator A5057913701 @default.
• W4224256927 date "2022-03-02" @default.
• W4224256927 modified "2023-09-26" @default.
• W4224256927 title "Computationally Efficient and Statistically Optimal Robust Low-rank Matrix and Tensor Estimation" @default.
• W4224256927 doi "https://doi.org/10.48550/arxiv.2203.00953" @default.
• W4224256927 hasPublicationYear "2022" @default.
• W4224256927 type Work @default.
• W4224256927 citedByCount "0" @default.
• W4224256927 crossrefType "posted-content" @default.
• W4224256927 hasAuthorship W4224256927A5034178721 @default.
• W4224256927 hasAuthorship W4224256927A5036610839 @default.
• W4224256927 hasAuthorship W4224256927A5043467180 @default.
• W4224256927 hasAuthorship W4224256927A5057913701 @default.
• W4224256927 hasBestOaLocation W42242569271 @default.
• W4224256927 hasConcept C105795698 @default.
• W4224256927 hasConcept C106487976 @default.
• W4224256927 hasConcept C112680207 @default.
• W4224256927 hasConcept C11413529 @default.
• W4224256927 hasConcept C114614502 @default.
• W4224256927 hasConcept C115961682 @default.
• W4224256927 hasConcept C118671147 @default.
• W4224256927 hasConcept C119857082 @default.
• W4224256927 hasConcept C121332964 @default.
• W4224256927 hasConcept C126255220 @default.
• W4224256927 hasConcept C127162648 @default.
• W4224256927 hasConcept C145446738 @default.
• W4224256927 hasConcept C153258448 @default.
• W4224256927 hasConcept C154945302 @default.
• W4224256927 hasConcept C159985019 @default.
• W4224256927 hasConcept C162324750 @default.
• W4224256927 hasConcept C163716315 @default.
• W4224256927 hasConcept C164226766 @default.
• W4224256927 hasConcept C185429906 @default.
• W4224256927 hasConcept C192562407 @default.
• W4224256927 hasConcept C206688291 @default.
• W4224256927 hasConcept C2524010 @default.
• W4224256927 hasConcept C2777303404 @default.
• W4224256927 hasConcept C2778459887 @default.
• W4224256927 hasConcept C28826006 @default.
• W4224256927 hasConcept C31258907 @default.
• W4224256927 hasConcept C33923547 @default.
• W4224256927 hasConcept C3770464 @default.
• W4224256927 hasConcept C41008148 @default.
• W4224256927 hasConcept C50522688 @default.
• W4224256927 hasConcept C50644808 @default.
• W4224256927 hasConcept C57869625 @default.
• W4224256927 hasConcept C62520636 @default.
• W4224256927 hasConcept C99498987 @default.
• W4224256927 hasConceptScore W4224256927C105795698 @default.
• W4224256927 hasConceptScore W4224256927C106487976 @default.
• W4224256927 hasConceptScore W4224256927C112680207 @default.
• W4224256927 hasConceptScore W4224256927C11413529 @default.
• W4224256927 hasConceptScore W4224256927C114614502 @default.
• W4224256927 hasConceptScore W4224256927C115961682 @default.
• W4224256927 hasConceptScore W4224256927C118671147 @default.
• W4224256927 hasConceptScore W4224256927C119857082 @default.
• W4224256927 hasConceptScore W4224256927C121332964 @default.
• W4224256927 hasConceptScore W4224256927C126255220 @default.
• W4224256927 hasConceptScore W4224256927C127162648 @default.
• W4224256927 hasConceptScore W4224256927C145446738 @default.
• W4224256927 hasConceptScore W4224256927C153258448 @default.
• W4224256927 hasConceptScore W4224256927C154945302 @default.
• W4224256927 hasConceptScore W4224256927C159985019 @default.
• W4224256927 hasConceptScore W4224256927C162324750 @default.
• W4224256927 hasConceptScore W4224256927C163716315 @default.
• W4224256927 hasConceptScore W4224256927C164226766 @default.
• W4224256927 hasConceptScore W4224256927C185429906 @default.
• W4224256927 hasConceptScore W4224256927C192562407 @default.
• W4224256927 hasConceptScore W4224256927C206688291 @default.
• W4224256927 hasConceptScore W4224256927C2524010 @default.
• W4224256927 hasConceptScore W4224256927C2777303404 @default.
• W4224256927 hasConceptScore W4224256927C2778459887 @default.
• W4224256927 hasConceptScore W4224256927C28826006 @default.
• W4224256927 hasConceptScore W4224256927C31258907 @default.
• W4224256927 hasConceptScore W4224256927C33923547 @default.
• W4224256927 hasConceptScore W4224256927C3770464 @default.
• W4224256927 hasConceptScore W4224256927C41008148 @default.
• W4224256927 hasConceptScore W4224256927C50522688 @default.
• W4224256927 hasConceptScore W4224256927C50644808 @default.
• W4224256927 hasConceptScore W4224256927C57869625 @default.
• W4224256927 hasConceptScore W4224256927C62520636 @default.
• W4224256927 hasConceptScore W4224256927C99498987 @default.
• W4224256927 hasLocation W42242569271 @default.
• W4224256927 hasOpenAccess W4224256927 @default.
• W4224256927 hasPrimaryLocation W42242569271 @default.
• W4224256927 hasRelatedWork W2078731137 @default.
• W4224256927 hasRelatedWork W2152704622 @default.
• W4224256927 hasRelatedWork W2267789472 @default.
• W4224256927 hasRelatedWork W2808489849 @default.
• W4224256927 hasRelatedWork W2928207644 @default.
• W4224256927 hasRelatedWork W2952134824 @default.
• W4224256927 hasRelatedWork W3081084973 @default.
• W4224256927 hasRelatedWork W4298055110 @default.
• W4224256927 hasRelatedWork W4301774174 @default.
• W4224256927 hasRelatedWork W4221144002 @default.

• next