Matches in SemOpenAlex for { <https://semopenalex.org/work/W3214334501> ?p ?o ?g. }
Showing items 1 to 76 of
76
with 100 items per page.
- W3214334501 abstract "Deep neural networks (DNNs) in the infinite width/channel limit have received much attention recently, as they provide a clear analytical window to deep learning via mappings to Gaussian Processes (GPs). Despite its theoretical appeal, this viewpoint lacks a crucial ingredient of deep learning in finite DNNs, laying at the heart of their success -- feature learning. Here we consider DNNs trained with noisy gradient descent on a large training set and derive a self consistent Gaussian Process theory accounting for strong finite-DNN and feature learning effects. Applying this to a toy model of a two-layer linear convolutional neural network (CNN) shows good agreement with experiments. We further identify, both analytical and numerically, a sharp transition between a feature learning regime and a lazy learning regime in this model. Strong finite-DNN effects are also derived for a non-linear two-layer fully connected network. Our self consistent theory provides a rich and versatile analytical framework for studying feature learning and other non-lazy effects in finite DNNs." @default.
- W3214334501 created "2021-11-22" @default.
- W3214334501 creator A5007159385 @default.
- W3214334501 creator A5053390682 @default.
- W3214334501 date "2021-12-06" @default.
- W3214334501 modified "2023-10-18" @default.
- W3214334501 title "A self consistent theory of Gaussian Processes captures feature learning effects in finite CNNs" @default.
- W3214334501 hasPublicationYear "2021" @default.
- W3214334501 type Work @default.
- W3214334501 sameAs 3214334501 @default.
- W3214334501 citedByCount "0" @default.
- W3214334501 crossrefType "proceedings-article" @default.
- W3214334501 hasAuthorship W3214334501A5007159385 @default.
- W3214334501 hasAuthorship W3214334501A5053390682 @default.
- W3214334501 hasConcept C108583219 @default.
- W3214334501 hasConcept C11413529 @default.
- W3214334501 hasConcept C121332964 @default.
- W3214334501 hasConcept C134306372 @default.
- W3214334501 hasConcept C138885662 @default.
- W3214334501 hasConcept C154945302 @default.
- W3214334501 hasConcept C162392398 @default.
- W3214334501 hasConcept C163716315 @default.
- W3214334501 hasConcept C2776401178 @default.
- W3214334501 hasConcept C33923547 @default.
- W3214334501 hasConcept C41008148 @default.
- W3214334501 hasConcept C41895202 @default.
- W3214334501 hasConcept C50644808 @default.
- W3214334501 hasConcept C59404180 @default.
- W3214334501 hasConcept C61326573 @default.
- W3214334501 hasConcept C62520636 @default.
- W3214334501 hasConcept C81363708 @default.
- W3214334501 hasConceptScore W3214334501C108583219 @default.
- W3214334501 hasConceptScore W3214334501C11413529 @default.
- W3214334501 hasConceptScore W3214334501C121332964 @default.
- W3214334501 hasConceptScore W3214334501C134306372 @default.
- W3214334501 hasConceptScore W3214334501C138885662 @default.
- W3214334501 hasConceptScore W3214334501C154945302 @default.
- W3214334501 hasConceptScore W3214334501C162392398 @default.
- W3214334501 hasConceptScore W3214334501C163716315 @default.
- W3214334501 hasConceptScore W3214334501C2776401178 @default.
- W3214334501 hasConceptScore W3214334501C33923547 @default.
- W3214334501 hasConceptScore W3214334501C41008148 @default.
- W3214334501 hasConceptScore W3214334501C41895202 @default.
- W3214334501 hasConceptScore W3214334501C50644808 @default.
- W3214334501 hasConceptScore W3214334501C59404180 @default.
- W3214334501 hasConceptScore W3214334501C61326573 @default.
- W3214334501 hasConceptScore W3214334501C62520636 @default.
- W3214334501 hasConceptScore W3214334501C81363708 @default.
- W3214334501 hasLocation W32143345011 @default.
- W3214334501 hasOpenAccess W3214334501 @default.
- W3214334501 hasPrimaryLocation W32143345011 @default.
- W3214334501 hasRelatedWork W2785807865 @default.
- W3214334501 hasRelatedWork W2799237514 @default.
- W3214334501 hasRelatedWork W2914483840 @default.
- W3214334501 hasRelatedWork W2927724204 @default.
- W3214334501 hasRelatedWork W2945236295 @default.
- W3214334501 hasRelatedWork W2956024278 @default.
- W3214334501 hasRelatedWork W2962685794 @default.
- W3214334501 hasRelatedWork W2963504252 @default.
- W3214334501 hasRelatedWork W2969554377 @default.
- W3214334501 hasRelatedWork W2970217468 @default.
- W3214334501 hasRelatedWork W2994720379 @default.
- W3214334501 hasRelatedWork W3013087590 @default.
- W3214334501 hasRelatedWork W3014864774 @default.
- W3214334501 hasRelatedWork W3084746729 @default.
- W3214334501 hasRelatedWork W3113190353 @default.
- W3214334501 hasRelatedWork W3157893192 @default.
- W3214334501 hasRelatedWork W3172130850 @default.
- W3214334501 hasRelatedWork W3175488619 @default.
- W3214334501 hasRelatedWork W3191734561 @default.
- W3214334501 hasRelatedWork W3200346102 @default.
- W3214334501 hasVolume "34" @default.
- W3214334501 isParatext "false" @default.
- W3214334501 isRetracted "false" @default.
- W3214334501 magId "3214334501" @default.
- W3214334501 workType "article" @default.