Matches in SemOpenAlex for { <https://semopenalex.org/work/W2970290137> ?p ?o ?g. }
Showing items 1 to 86 of
86
with 100 items per page.
- W2970290137 endingPage "11622" @default.
- W2970290137 startingPage "11611" @default.
- W2970290137 abstract "Aimed at explaining the surprisingly good generalization behavior of overparameterized deep networks, recent works have developed a variety of generalization bounds for deep learning, all based on the fundamental learning-theoretic technique of uniform convergence. While it is well-known that many of these existing bounds are numerically large, through numerous experiments, we bring to light a more concerning aspect of these bounds: in practice, these bounds can {em increase} with the training dataset size. Guided by our observations, we then present examples of overparameterized linear classifiers and neural networks trained by gradient descent (GD) where uniform convergence provably cannot ``explain generalization'' -- even if we take into account the implicit bias of GD {em to the fullest extent possible}. More precisely, even if we consider only the set of classifiers output by GD, which have test errors less than some small $epsilon$ in our settings, we show that applying (two-sided) uniform convergence on this set of classifiers will yield only a vacuous generalization guarantee larger than $1-epsilon$. Through these findings, we cast doubt on the power of uniform convergence-based generalization bounds to provide a complete picture of why overparameterized deep networks generalize well." @default.
- W2970290137 created "2019-09-05" @default.
- W2970290137 creator A5051270969 @default.
- W2970290137 creator A5075035644 @default.
- W2970290137 date "2019-09-06" @default.
- W2970290137 modified "2023-09-24" @default.
- W2970290137 title "Uniform convergence may be unable to explain generalization in deep learning" @default.
- W2970290137 hasPublicationYear "2019" @default.
- W2970290137 type Work @default.
- W2970290137 sameAs 2970290137 @default.
- W2970290137 citedByCount "91" @default.
- W2970290137 countsByYear W29702901372019 @default.
- W2970290137 countsByYear W29702901372020 @default.
- W2970290137 countsByYear W29702901372021 @default.
- W2970290137 countsByYear W29702901372022 @default.
- W2970290137 crossrefType "proceedings-article" @default.
- W2970290137 hasAuthorship W2970290137A5051270969 @default.
- W2970290137 hasAuthorship W2970290137A5075035644 @default.
- W2970290137 hasConcept C108583219 @default.
- W2970290137 hasConcept C11413529 @default.
- W2970290137 hasConcept C119857082 @default.
- W2970290137 hasConcept C134306372 @default.
- W2970290137 hasConcept C153258448 @default.
- W2970290137 hasConcept C154945302 @default.
- W2970290137 hasConcept C162324750 @default.
- W2970290137 hasConcept C169903167 @default.
- W2970290137 hasConcept C177148314 @default.
- W2970290137 hasConcept C177264268 @default.
- W2970290137 hasConcept C199360897 @default.
- W2970290137 hasConcept C2777303404 @default.
- W2970290137 hasConcept C28826006 @default.
- W2970290137 hasConcept C2984842247 @default.
- W2970290137 hasConcept C33923547 @default.
- W2970290137 hasConcept C41008148 @default.
- W2970290137 hasConcept C50522688 @default.
- W2970290137 hasConcept C50644808 @default.
- W2970290137 hasConcept C5465570 @default.
- W2970290137 hasConceptScore W2970290137C108583219 @default.
- W2970290137 hasConceptScore W2970290137C11413529 @default.
- W2970290137 hasConceptScore W2970290137C119857082 @default.
- W2970290137 hasConceptScore W2970290137C134306372 @default.
- W2970290137 hasConceptScore W2970290137C153258448 @default.
- W2970290137 hasConceptScore W2970290137C154945302 @default.
- W2970290137 hasConceptScore W2970290137C162324750 @default.
- W2970290137 hasConceptScore W2970290137C169903167 @default.
- W2970290137 hasConceptScore W2970290137C177148314 @default.
- W2970290137 hasConceptScore W2970290137C177264268 @default.
- W2970290137 hasConceptScore W2970290137C199360897 @default.
- W2970290137 hasConceptScore W2970290137C2777303404 @default.
- W2970290137 hasConceptScore W2970290137C28826006 @default.
- W2970290137 hasConceptScore W2970290137C2984842247 @default.
- W2970290137 hasConceptScore W2970290137C33923547 @default.
- W2970290137 hasConceptScore W2970290137C41008148 @default.
- W2970290137 hasConceptScore W2970290137C50522688 @default.
- W2970290137 hasConceptScore W2970290137C50644808 @default.
- W2970290137 hasConceptScore W2970290137C5465570 @default.
- W2970290137 hasLocation W29702901371 @default.
- W2970290137 hasOpenAccess W2970290137 @default.
- W2970290137 hasPrimaryLocation W29702901371 @default.
- W2970290137 hasRelatedWork W2014384147 @default.
- W2970290137 hasRelatedWork W2139338362 @default.
- W2970290137 hasRelatedWork W2194775991 @default.
- W2970290137 hasRelatedWork W2579923771 @default.
- W2970290137 hasRelatedWork W2604117713 @default.
- W2970290137 hasRelatedWork W2809090039 @default.
- W2970290137 hasRelatedWork W2911742574 @default.
- W2970290137 hasRelatedWork W2914852400 @default.
- W2970290137 hasRelatedWork W2923764619 @default.
- W2970290137 hasRelatedWork W2962857907 @default.
- W2970290137 hasRelatedWork W2963236897 @default.
- W2970290137 hasRelatedWork W2963285844 @default.
- W2970290137 hasRelatedWork W2963518130 @default.
- W2970290137 hasRelatedWork W2963664410 @default.
- W2970290137 hasRelatedWork W2963695615 @default.
- W2970290137 hasRelatedWork W2963739978 @default.
- W2970290137 hasRelatedWork W3118608800 @default.
- W2970290137 hasRelatedWork W3119586787 @default.
- W2970290137 hasRelatedWork W3137695714 @default.
- W2970290137 hasRelatedWork W607505555 @default.
- W2970290137 hasVolume "32" @default.
- W2970290137 isParatext "false" @default.
- W2970290137 isRetracted "false" @default.
- W2970290137 magId "2970290137" @default.
- W2970290137 workType "article" @default.