Matches in SemOpenAlex for { <https://semopenalex.org/work/W3033629543> ?p ?o ?g. }
- W3033629543 abstract "In 1988, Eric B. Baum showed that two-layers neural networks with threshold activation function can perfectly memorize the binary labels of $n$ points in general position in $mathbb{R}^d$ using only $ulcorner n/d urcorner$ neurons. We observe that with ReLU networks, using four times as many neurons one can fit arbitrary real labels. Moreover, for approximate memorization up to error $epsilon$, the neural tangent kernel can also memorize with only $Oleft(frac{n}{d} cdot log(1/epsilon) right)$ neurons (assuming that the data is well dispersed too). We show however that these constructions give rise to networks where the magnitude of the neurons' weights are far from optimal. In contrast we propose a new training procedure for ReLU networks, based on complex (as opposed to real) recombination of the neurons, for which we show approximate memorization with both $Oleft(frac{n}{d} cdot frac{log(1/epsilon)}{epsilon}right)$ neurons, as well as nearly-optimal size of the weights." @default.
- W3033629543 created "2020-06-12" @default.
- W3033629543 creator A5011758954 @default.
- W3033629543 creator A5021842384 @default.
- W3033629543 creator A5024980585 @default.
- W3033629543 creator A5086897744 @default.
- W3033629543 date "2020-06-04" @default.
- W3033629543 modified "2023-09-27" @default.
- W3033629543 title "Network size and weights size for memorization with two-layers neural networks" @default.
- W3033629543 cites W2012903341 @default.
- W3033629543 cites W2017834428 @default.
- W3033629543 cites W2099579348 @default.
- W3033629543 cites W2103496339 @default.
- W3033629543 cites W2106458073 @default.
- W3033629543 cites W2113517874 @default.
- W3033629543 cites W2166116275 @default.
- W3033629543 cites W2167967601 @default.
- W3033629543 cites W2809090039 @default.
- W3033629543 cites W2886067286 @default.
- W3033629543 cites W2913892099 @default.
- W3033629543 cites W2952204734 @default.
- W3033629543 cites W2962698540 @default.
- W3033629543 cites W2962857907 @default.
- W3033629543 cites W2963095610 @default.
- W3033629543 cites W2963239103 @default.
- W3033629543 cites W2963417959 @default.
- W3033629543 cites W2963534251 @default.
- W3033629543 cites W2963664410 @default.
- W3033629543 cites W2964098911 @default.
- W3033629543 cites W2964624822 @default.
- W3033629543 cites W2970618525 @default.
- W3033629543 cites W2970723196 @default.
- W3033629543 cites W2971120029 @default.
- W3033629543 cites W2990393274 @default.
- W3033629543 cites W2993367001 @default.
- W3033629543 cites W2995354826 @default.
- W3033629543 cites W2996168800 @default.
- W3033629543 cites W3003477032 @default.
- W3033629543 cites W3013980252 @default.
- W3033629543 cites W3125537303 @default.
- W3033629543 cites W3137695714 @default.
- W3033629543 cites W607505555 @default.
- W3033629543 hasPublicationYear "2020" @default.
- W3033629543 type Work @default.
- W3033629543 sameAs 3033629543 @default.
- W3033629543 citedByCount "15" @default.
- W3033629543 countsByYear W30336295432020 @default.
- W3033629543 countsByYear W30336295432021 @default.
- W3033629543 crossrefType "posted-content" @default.
- W3033629543 hasAuthorship W3033629543A5011758954 @default.
- W3033629543 hasAuthorship W3033629543A5021842384 @default.
- W3033629543 hasAuthorship W3033629543A5024980585 @default.
- W3033629543 hasAuthorship W3033629543A5086897744 @default.
- W3033629543 hasConcept C11413529 @default.
- W3033629543 hasConcept C114614502 @default.
- W3033629543 hasConcept C118615104 @default.
- W3033629543 hasConcept C121332964 @default.
- W3033629543 hasConcept C14036430 @default.
- W3033629543 hasConcept C145420912 @default.
- W3033629543 hasConcept C154945302 @default.
- W3033629543 hasConcept C199360897 @default.
- W3033629543 hasConcept C30038468 @default.
- W3033629543 hasConcept C33923547 @default.
- W3033629543 hasConcept C41008148 @default.
- W3033629543 hasConcept C48372109 @default.
- W3033629543 hasConcept C50644808 @default.
- W3033629543 hasConcept C74193536 @default.
- W3033629543 hasConcept C78458016 @default.
- W3033629543 hasConcept C86803240 @default.
- W3033629543 hasConcept C94375191 @default.
- W3033629543 hasConcept C97137487 @default.
- W3033629543 hasConceptScore W3033629543C11413529 @default.
- W3033629543 hasConceptScore W3033629543C114614502 @default.
- W3033629543 hasConceptScore W3033629543C118615104 @default.
- W3033629543 hasConceptScore W3033629543C121332964 @default.
- W3033629543 hasConceptScore W3033629543C14036430 @default.
- W3033629543 hasConceptScore W3033629543C145420912 @default.
- W3033629543 hasConceptScore W3033629543C154945302 @default.
- W3033629543 hasConceptScore W3033629543C199360897 @default.
- W3033629543 hasConceptScore W3033629543C30038468 @default.
- W3033629543 hasConceptScore W3033629543C33923547 @default.
- W3033629543 hasConceptScore W3033629543C41008148 @default.
- W3033629543 hasConceptScore W3033629543C48372109 @default.
- W3033629543 hasConceptScore W3033629543C50644808 @default.
- W3033629543 hasConceptScore W3033629543C74193536 @default.
- W3033629543 hasConceptScore W3033629543C78458016 @default.
- W3033629543 hasConceptScore W3033629543C86803240 @default.
- W3033629543 hasConceptScore W3033629543C94375191 @default.
- W3033629543 hasConceptScore W3033629543C97137487 @default.
- W3033629543 hasLocation W30336295431 @default.
- W3033629543 hasOpenAccess W3033629543 @default.
- W3033629543 hasPrimaryLocation W30336295431 @default.
- W3033629543 hasRelatedWork W1986477811 @default.
- W3033629543 hasRelatedWork W2012903341 @default.
- W3033629543 hasRelatedWork W2103496339 @default.
- W3033629543 hasRelatedWork W2158368251 @default.
- W3033629543 hasRelatedWork W2221366174 @default.
- W3033629543 hasRelatedWork W2557635520 @default.
- W3033629543 hasRelatedWork W2809090039 @default.
- W3033629543 hasRelatedWork W2951548721 @default.