Matches in SemOpenAlex for { <https://semopenalex.org/work/W2787442904> ?p ?o ?g. }
Showing items 1 to 83 of
83
with 100 items per page.
- W2787442904 abstract "The recent success of deep neural networks stems from their ability to generalize well on real data; however, Zhang et al. have observed that neural networks can easily overfit random labels. This observation demonstrates that with the existing theory, we cannot adequately explain why gradient methods can find generalizable solutions for neural networks. In this work, we use a Fourier-based approach to study the generalization properties of gradient-based methods over 2-layer neural networks with sinusoidal activation functions. We prove that if the underlying distribution of data has nice spectral properties such as bandlimitedness, then the gradient descent method will converge to generalizable local minima. We also establish a Fourier-based generalization bound for bandlimited spaces, which generalizes to other activation functions. Our generalization bound motivates a grouped version of path norms for measuring the complexity of 2-layer neural networks with ReLU activation functions. We demonstrate numerically that regularization of this group path norm results in neural network solutions that can fit true labels without losing test accuracy while not overfitting random labels." @default.
- W2787442904 created "2018-02-23" @default.
- W2787442904 creator A5011123850 @default.
- W2787442904 creator A5017160178 @default.
- W2787442904 creator A5024072566 @default.
- W2787442904 date "2018-02-15" @default.
- W2787442904 modified "2023-09-26" @default.
- W2787442904 title "A Spectral Approach to Generalization and Optimization in Neural Networks" @default.
- W2787442904 hasPublicationYear "2018" @default.
- W2787442904 type Work @default.
- W2787442904 sameAs 2787442904 @default.
- W2787442904 citedByCount "3" @default.
- W2787442904 countsByYear W27874429042019 @default.
- W2787442904 countsByYear W27874429042020 @default.
- W2787442904 crossrefType "journal-article" @default.
- W2787442904 hasAuthorship W2787442904A5011123850 @default.
- W2787442904 hasAuthorship W2787442904A5017160178 @default.
- W2787442904 hasAuthorship W2787442904A5024072566 @default.
- W2787442904 hasConcept C102519508 @default.
- W2787442904 hasConcept C11413529 @default.
- W2787442904 hasConcept C129997835 @default.
- W2787442904 hasConcept C134306372 @default.
- W2787442904 hasConcept C153258448 @default.
- W2787442904 hasConcept C154945302 @default.
- W2787442904 hasConcept C175202392 @default.
- W2787442904 hasConcept C177148314 @default.
- W2787442904 hasConcept C186633575 @default.
- W2787442904 hasConcept C199360897 @default.
- W2787442904 hasConcept C206688291 @default.
- W2787442904 hasConcept C22019652 @default.
- W2787442904 hasConcept C2776135515 @default.
- W2787442904 hasConcept C2777735758 @default.
- W2787442904 hasConcept C33923547 @default.
- W2787442904 hasConcept C38365724 @default.
- W2787442904 hasConcept C41008148 @default.
- W2787442904 hasConcept C50644808 @default.
- W2787442904 hasConcept C86582703 @default.
- W2787442904 hasConceptScore W2787442904C102519508 @default.
- W2787442904 hasConceptScore W2787442904C11413529 @default.
- W2787442904 hasConceptScore W2787442904C129997835 @default.
- W2787442904 hasConceptScore W2787442904C134306372 @default.
- W2787442904 hasConceptScore W2787442904C153258448 @default.
- W2787442904 hasConceptScore W2787442904C154945302 @default.
- W2787442904 hasConceptScore W2787442904C175202392 @default.
- W2787442904 hasConceptScore W2787442904C177148314 @default.
- W2787442904 hasConceptScore W2787442904C186633575 @default.
- W2787442904 hasConceptScore W2787442904C199360897 @default.
- W2787442904 hasConceptScore W2787442904C206688291 @default.
- W2787442904 hasConceptScore W2787442904C22019652 @default.
- W2787442904 hasConceptScore W2787442904C2776135515 @default.
- W2787442904 hasConceptScore W2787442904C2777735758 @default.
- W2787442904 hasConceptScore W2787442904C33923547 @default.
- W2787442904 hasConceptScore W2787442904C38365724 @default.
- W2787442904 hasConceptScore W2787442904C41008148 @default.
- W2787442904 hasConceptScore W2787442904C50644808 @default.
- W2787442904 hasConceptScore W2787442904C86582703 @default.
- W2787442904 hasLocation W27874429041 @default.
- W2787442904 hasOpenAccess W2787442904 @default.
- W2787442904 hasPrimaryLocation W27874429041 @default.
- W2787442904 hasRelatedWork W2897080984 @default.
- W2787442904 hasRelatedWork W2913473169 @default.
- W2787442904 hasRelatedWork W2919767206 @default.
- W2787442904 hasRelatedWork W2927724204 @default.
- W2787442904 hasRelatedWork W2944820194 @default.
- W2787442904 hasRelatedWork W2980702925 @default.
- W2787442904 hasRelatedWork W2995015865 @default.
- W2787442904 hasRelatedWork W3014013215 @default.
- W2787442904 hasRelatedWork W3034649595 @default.
- W2787442904 hasRelatedWork W3039204554 @default.
- W2787442904 hasRelatedWork W3095034705 @default.
- W2787442904 hasRelatedWork W3096184750 @default.
- W2787442904 hasRelatedWork W3108216709 @default.
- W2787442904 hasRelatedWork W3153348027 @default.
- W2787442904 hasRelatedWork W3162003518 @default.
- W2787442904 hasRelatedWork W3192896295 @default.
- W2787442904 hasRelatedWork W3193396833 @default.
- W2787442904 hasRelatedWork W3199400829 @default.
- W2787442904 hasRelatedWork W3206818904 @default.
- W2787442904 hasRelatedWork W3208268610 @default.
- W2787442904 isParatext "false" @default.
- W2787442904 isRetracted "false" @default.
- W2787442904 magId "2787442904" @default.
- W2787442904 workType "article" @default.