Matches in SemOpenAlex for { <https://semopenalex.org/work/W2970638103> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W2970638103 endingPage "11023" @default.
- W2970638103 startingPage "11013" @default.
- W2970638103 abstract "In this work, we improve upon the stepwise analysis of noisy iterative learning algorithms initiated by Pensia, Jog, and Loh (2018) and recently extended by Bu, Zou, and Veeravalli (2019). Our main contributions are significantly improved mutual information bounds for Stochastic Gradient Langevin Dynamics via data-dependent estimates. Our approach is based on the variational characterization of mutual information and the use of data-dependent priors that forecast the mini-batch gradient based on a subset of the training samples. Our approach is broadly applicable within the information-theoretic framework of Russo and Zou (2015) and Xu and Raginsky (2017). Our bound can be tied to a measure of flatness of the empirical risk surface. As compared with other bounds that depend on the squared norms of gradients, empirical investigations show that the terms in our bounds are orders of magnitude smaller." @default.
- W2970638103 created "2019-09-05" @default.
- W2970638103 creator A5019870310 @default.
- W2970638103 creator A5021062973 @default.
- W2970638103 creator A5042384857 @default.
- W2970638103 creator A5046253142 @default.
- W2970638103 creator A5061900418 @default.
- W2970638103 date "2019-01-01" @default.
- W2970638103 modified "2023-10-18" @default.
- W2970638103 title "Information-Theoretic Generalization Bounds for SGLD via Data-Dependent Estimates" @default.
- W2970638103 hasPublicationYear "2019" @default.
- W2970638103 type Work @default.
- W2970638103 sameAs 2970638103 @default.
- W2970638103 citedByCount "32" @default.
- W2970638103 countsByYear W29706381032019 @default.
- W2970638103 countsByYear W29706381032020 @default.
- W2970638103 countsByYear W29706381032021 @default.
- W2970638103 crossrefType "proceedings-article" @default.
- W2970638103 hasAuthorship W2970638103A5019870310 @default.
- W2970638103 hasAuthorship W2970638103A5021062973 @default.
- W2970638103 hasAuthorship W2970638103A5042384857 @default.
- W2970638103 hasAuthorship W2970638103A5046253142 @default.
- W2970638103 hasAuthorship W2970638103A5061900418 @default.
- W2970638103 hasConcept C105795698 @default.
- W2970638103 hasConcept C124101348 @default.
- W2970638103 hasConcept C126255220 @default.
- W2970638103 hasConcept C134306372 @default.
- W2970638103 hasConcept C152139883 @default.
- W2970638103 hasConcept C177148314 @default.
- W2970638103 hasConcept C2780004032 @default.
- W2970638103 hasConcept C2780009758 @default.
- W2970638103 hasConcept C28826006 @default.
- W2970638103 hasConcept C33923547 @default.
- W2970638103 hasConcept C41008148 @default.
- W2970638103 hasConcept C77553402 @default.
- W2970638103 hasConceptScore W2970638103C105795698 @default.
- W2970638103 hasConceptScore W2970638103C124101348 @default.
- W2970638103 hasConceptScore W2970638103C126255220 @default.
- W2970638103 hasConceptScore W2970638103C134306372 @default.
- W2970638103 hasConceptScore W2970638103C152139883 @default.
- W2970638103 hasConceptScore W2970638103C177148314 @default.
- W2970638103 hasConceptScore W2970638103C2780004032 @default.
- W2970638103 hasConceptScore W2970638103C2780009758 @default.
- W2970638103 hasConceptScore W2970638103C28826006 @default.
- W2970638103 hasConceptScore W2970638103C33923547 @default.
- W2970638103 hasConceptScore W2970638103C41008148 @default.
- W2970638103 hasConceptScore W2970638103C77553402 @default.
- W2970638103 hasLocation W29706381031 @default.
- W2970638103 hasOpenAccess W2970638103 @default.
- W2970638103 hasPrimaryLocation W29706381031 @default.
- W2970638103 hasRelatedWork W2014384147 @default.
- W2970638103 hasRelatedWork W2121358800 @default.
- W2970638103 hasRelatedWork W2139338362 @default.
- W2970638103 hasRelatedWork W2167433878 @default.
- W2970638103 hasRelatedWork W2258658829 @default.
- W2970638103 hasRelatedWork W2567352378 @default.
- W2970638103 hasRelatedWork W2910572885 @default.
- W2970638103 hasRelatedWork W2962702650 @default.
- W2970638103 hasRelatedWork W2963650125 @default.
- W2970638103 hasRelatedWork W2963862692 @default.
- W2970638103 hasRelatedWork W2979452771 @default.
- W2970638103 hasRelatedWork W3019669224 @default.
- W2970638103 hasRelatedWork W3024982789 @default.
- W2970638103 hasRelatedWork W3036393886 @default.
- W2970638103 hasRelatedWork W3046489903 @default.
- W2970638103 hasRelatedWork W3100231902 @default.
- W2970638103 hasRelatedWork W3104240813 @default.
- W2970638103 hasRelatedWork W3137695714 @default.
- W2970638103 hasRelatedWork W568673721 @default.
- W2970638103 hasRelatedWork W607505555 @default.
- W2970638103 hasVolume "32" @default.
- W2970638103 isParatext "false" @default.
- W2970638103 isRetracted "false" @default.
- W2970638103 magId "2970638103" @default.
- W2970638103 workType "article" @default.