FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2914671377> ?p ?o ?g. }

• W2914671377 abstract "Generalization error (also known as the out-of-sample error) measures how well the hypothesis learned from training data generalizes to previously unseen data. Proving tight generalization error bounds is a central question in statistical learning theory. In this paper, we obtain generalization error bounds for learning general non-convex objectives, which has attracted significant attention in recent years. We develop a new framework, termed Bayes-Stability, for proving algorithm-dependent generalization error bounds. The new framework combines ideas from both the PAC-Bayesian theory and the notion of algorithmic stability. Applying the Bayes-Stability method, we obtain new data-dependent generalization bounds for stochastic gradient Langevin dynamics (SGLD) and several other noisy gradient methods (e.g., with momentum, mini-batch and acceleration, Entropy-SGD). Our result recovers (and is typically tighter than) a recent result in Mou et al. (2018) and improves upon the results in Pensia et al. (2018). Our experiments demonstrate that our data-dependent bounds can distinguish randomly labelled data from normal data, which provides an explanation to the intriguing phenomena observed in Zhang et al. (2017a). We also study the setting where the total loss is the sum of a bounded loss and an additional ell_2 regularization term. We obtain new generalization bounds for the continuous Langevin dynamic in this setting by developing a new Log-Sobolev inequality for the parameter distribution at any time. Our new bounds are more desirable when the noisy level of the process is not small, and do not become vacuous even when T tends to infinity." @default.
• W2914671377 created "2019-02-21" @default.
• W2914671377 creator A5023118987 @default.
• W2914671377 creator A5052883326 @default.
• W2914671377 creator A5067257960 @default.
• W2914671377 date "2019-02-01" @default.
• W2914671377 modified "2023-09-23" @default.
• W2914671377 title "On Generalization Error Bounds of Noisy Gradient Methods for Non-Convex Learning" @default.
• W2914671377 cites W104184427 @default.
• W2914671377 cites W1575877217 @default.
• W2914671377 cites W1592998450 @default.
• W2914671377 cites W1974932891 @default.
• W2914671377 cites W1988720110 @default.
• W2914671377 cites W1999376700 @default.
• W2914671377 cites W2046392353 @default.
• W2914671377 cites W2112796928 @default.
• W2914671377 cites W2131542408 @default.
• W2914671377 cites W2139338362 @default.
• W2914671377 cites W2159902206 @default.
• W2914671377 cites W2163605009 @default.
• W2914671377 cites W2167433878 @default.
• W2914671377 cites W2323824333 @default.
• W2914671377 cites W2604451472 @default.
• W2914671377 cites W2795605442 @default.
• W2914671377 cites W2900959181 @default.
• W2914671377 cites W2946963372 @default.
• W2914671377 cites W2952204734 @default.
• W2914671377 cites W2962698540 @default.
• W2914671377 cites W2962702650 @default.
• W2914671377 cites W2962839956 @default.
• W2914671377 cites W2962857907 @default.
• W2914671377 cites W2963145271 @default.
• W2914671377 cites W2963236897 @default.
• W2914671377 cites W2963285844 @default.
• W2914671377 cites W2963375484 @default.
• W2914671377 cites W2963481221 @default.
• W2914671377 cites W2963509076 @default.
• W2914671377 cites W2963532386 @default.
• W2914671377 cites W2963534466 @default.
• W2914671377 cites W2963664410 @default.
• W2914671377 cites W2963794891 @default.
• W2914671377 cites W2963862692 @default.
• W2914671377 cites W2963874210 @default.
• W2914671377 cites W2964163109 @default.
• W2914671377 cites W2965157832 @default.
• W2914671377 cites W2970330753 @default.
• W2914671377 cites W2971043187 @default.
• W2914671377 cites W2971069898 @default.
• W2914671377 cites W3093329015 @default.
• W2914671377 cites W3112350859 @default.
• W2914671377 cites W3118608800 @default.
• W2914671377 cites W3137695714 @default.
• W2914671377 cites W423603040 @default.
• W2914671377 doi "https://doi.org/10.48550/arxiv.1902.00621" @default.
• W2914671377 hasPublicationYear "2019" @default.
• W2914671377 type Work @default.
• W2914671377 sameAs 2914671377 @default.
• W2914671377 citedByCount "17" @default.
• W2914671377 countsByYear W29146713772019 @default.
• W2914671377 countsByYear W29146713772020 @default.
• W2914671377 countsByYear W29146713772021 @default.
• W2914671377 crossrefType "posted-content" @default.
• W2914671377 hasAuthorship W2914671377A5023118987 @default.
• W2914671377 hasAuthorship W2914671377A5052883326 @default.
• W2914671377 hasAuthorship W2914671377A5067257960 @default.
• W2914671377 hasBestOaLocation W29146713771 @default.
• W2914671377 hasConcept C105795698 @default.
• W2914671377 hasConcept C112680207 @default.
• W2914671377 hasConcept C112972136 @default.
• W2914671377 hasConcept C11413529 @default.
• W2914671377 hasConcept C119857082 @default.
• W2914671377 hasConcept C134306372 @default.
• W2914671377 hasConcept C154945302 @default.
• W2914671377 hasConcept C177148314 @default.
• W2914671377 hasConcept C17744445 @default.
• W2914671377 hasConcept C191795146 @default.
• W2914671377 hasConcept C199539241 @default.
• W2914671377 hasConcept C2524010 @default.
• W2914671377 hasConcept C2776135515 @default.
• W2914671377 hasConcept C2780004032 @default.
• W2914671377 hasConcept C28826006 @default.
• W2914671377 hasConcept C33923547 @default.
• W2914671377 hasConcept C34388435 @default.
• W2914671377 hasConcept C41008148 @default.
• W2914671377 hasConceptScore W2914671377C105795698 @default.
• W2914671377 hasConceptScore W2914671377C112680207 @default.
• W2914671377 hasConceptScore W2914671377C112972136 @default.
• W2914671377 hasConceptScore W2914671377C11413529 @default.
• W2914671377 hasConceptScore W2914671377C119857082 @default.
• W2914671377 hasConceptScore W2914671377C134306372 @default.
• W2914671377 hasConceptScore W2914671377C154945302 @default.
• W2914671377 hasConceptScore W2914671377C177148314 @default.
• W2914671377 hasConceptScore W2914671377C17744445 @default.
• W2914671377 hasConceptScore W2914671377C191795146 @default.
• W2914671377 hasConceptScore W2914671377C199539241 @default.
• W2914671377 hasConceptScore W2914671377C2524010 @default.
• W2914671377 hasConceptScore W2914671377C2776135515 @default.
• W2914671377 hasConceptScore W2914671377C2780004032 @default.
• W2914671377 hasConceptScore W2914671377C28826006 @default.
• W2914671377 hasConceptScore W2914671377C33923547 @default.

• next