Matches in SemOpenAlex for { <https://semopenalex.org/work/W3034851139> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W3034851139 endingPage "5819" @default.
- W3034851139 startingPage "5809" @default.
- W3034851139 abstract "Recently there are a considerable amount of work devoted to the study of the algorithmic stability and generalization for stochastic gradient descent (SGD). However, the existing stability analysis requires to impose restrictive assumptions on the boundedness of gradients, strong smoothness and convexity of loss functions. In this paper, we provide a fine-grained analysis of stability and generalization for SGD by substantially relaxing these assumptions. Firstly, we establish stability and generalization for SGD by removing the existing bounded gradient assumptions. The key idea is the introduction of a new stability measure called on-average model stability, for which we develop novel bounds controlled by the risks of SGD iterates. This yields generalization bounds depending on the behavior of the best model, and leads to the first-ever-known fast bounds in the low-noise setting using stability approach. Secondly, the smoothness assumption is relaxed by considering loss functions with Holder continuous (sub)gradients for which we show that optimal bounds are still achieved by balancing computation and stability. To our best knowledge, this gives the first-ever-known stability and generalization bounds for SGD with even non-differentiable loss functions. Finally, we study learning problems with (strongly) convex objectives but non-convex loss functions." @default.
- W3034851139 created "2020-06-19" @default.
- W3034851139 creator A5046468616 @default.
- W3034851139 creator A5048960543 @default.
- W3034851139 date "2020-07-12" @default.
- W3034851139 modified "2023-09-22" @default.
- W3034851139 title "Fine-Grained Analysis of Stability and Generalization for Stochastic Gradient Descent" @default.
- W3034851139 hasPublicationYear "2020" @default.
- W3034851139 type Work @default.
- W3034851139 sameAs 3034851139 @default.
- W3034851139 citedByCount "14" @default.
- W3034851139 countsByYear W30348511392020 @default.
- W3034851139 countsByYear W30348511392021 @default.
- W3034851139 crossrefType "proceedings-article" @default.
- W3034851139 hasAuthorship W3034851139A5046468616 @default.
- W3034851139 hasAuthorship W3034851139A5048960543 @default.
- W3034851139 hasConcept C102634674 @default.
- W3034851139 hasConcept C106159729 @default.
- W3034851139 hasConcept C112680207 @default.
- W3034851139 hasConcept C112972136 @default.
- W3034851139 hasConcept C119857082 @default.
- W3034851139 hasConcept C126255220 @default.
- W3034851139 hasConcept C134306372 @default.
- W3034851139 hasConcept C145446738 @default.
- W3034851139 hasConcept C153258448 @default.
- W3034851139 hasConcept C154945302 @default.
- W3034851139 hasConcept C162324750 @default.
- W3034851139 hasConcept C177148314 @default.
- W3034851139 hasConcept C202615002 @default.
- W3034851139 hasConcept C206688291 @default.
- W3034851139 hasConcept C2524010 @default.
- W3034851139 hasConcept C28826006 @default.
- W3034851139 hasConcept C33923547 @default.
- W3034851139 hasConcept C34388435 @default.
- W3034851139 hasConcept C41008148 @default.
- W3034851139 hasConcept C50644808 @default.
- W3034851139 hasConcept C72134830 @default.
- W3034851139 hasConcept C77553402 @default.
- W3034851139 hasConceptScore W3034851139C102634674 @default.
- W3034851139 hasConceptScore W3034851139C106159729 @default.
- W3034851139 hasConceptScore W3034851139C112680207 @default.
- W3034851139 hasConceptScore W3034851139C112972136 @default.
- W3034851139 hasConceptScore W3034851139C119857082 @default.
- W3034851139 hasConceptScore W3034851139C126255220 @default.
- W3034851139 hasConceptScore W3034851139C134306372 @default.
- W3034851139 hasConceptScore W3034851139C145446738 @default.
- W3034851139 hasConceptScore W3034851139C153258448 @default.
- W3034851139 hasConceptScore W3034851139C154945302 @default.
- W3034851139 hasConceptScore W3034851139C162324750 @default.
- W3034851139 hasConceptScore W3034851139C177148314 @default.
- W3034851139 hasConceptScore W3034851139C202615002 @default.
- W3034851139 hasConceptScore W3034851139C206688291 @default.
- W3034851139 hasConceptScore W3034851139C2524010 @default.
- W3034851139 hasConceptScore W3034851139C28826006 @default.
- W3034851139 hasConceptScore W3034851139C33923547 @default.
- W3034851139 hasConceptScore W3034851139C34388435 @default.
- W3034851139 hasConceptScore W3034851139C41008148 @default.
- W3034851139 hasConceptScore W3034851139C50644808 @default.
- W3034851139 hasConceptScore W3034851139C72134830 @default.
- W3034851139 hasConceptScore W3034851139C77553402 @default.
- W3034851139 hasLocation W30348511391 @default.
- W3034851139 hasOpenAccess W3034851139 @default.
- W3034851139 hasPrimaryLocation W30348511391 @default.
- W3034851139 hasRelatedWork W1992208280 @default.
- W3034851139 hasRelatedWork W2097931325 @default.
- W3034851139 hasRelatedWork W2113651538 @default.
- W3034851139 hasRelatedWork W2131542408 @default.
- W3034851139 hasRelatedWork W2139338362 @default.
- W3034851139 hasRelatedWork W2604451472 @default.
- W3034851139 hasRelatedWork W27434444 @default.
- W3034851139 hasRelatedWork W2788714855 @default.
- W3034851139 hasRelatedWork W2795605442 @default.
- W3034851139 hasRelatedWork W2963094221 @default.
- W3034851139 hasRelatedWork W2963248893 @default.
- W3034851139 hasRelatedWork W2963794891 @default.
- W3034851139 hasRelatedWork W2965157832 @default.
- W3034851139 hasRelatedWork W2994760360 @default.
- W3034851139 hasRelatedWork W3034324151 @default.
- W3034851139 hasRelatedWork W3046508829 @default.
- W3034851139 hasRelatedWork W3098458959 @default.
- W3034851139 hasRelatedWork W3150569225 @default.
- W3034851139 hasRelatedWork W3169267543 @default.
- W3034851139 hasRelatedWork W607505555 @default.
- W3034851139 isParatext "false" @default.
- W3034851139 isRetracted "false" @default.
- W3034851139 magId "3034851139" @default.
- W3034851139 workType "article" @default.