FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2899297079> ?p ?o ?g. }

• W2899297079 endingPage "82" @default.
• W2899297079 startingPage "31" @default.
• W2899297079 abstract "We investigated the feature map inside deep neural networks (DNNs) by tracking the transport map. We are interested in the role of depth--why do DNNs perform better than shallow models?--and the interpretation of DNNs--what do intermediate layers do? Despite the rapid development in their application, DNNs remain analytically unexplained because the hidden layers are nested and the parameters are not faithful. Inspired by the integral representation of shallow NNs, which is the continuum limit of the width, or the hidden unit number, we developed the flow representation and transport analysis of DNNs. The flow representation is the continuum limit of the depth, or the hidden layer number, and it is specified by an ordinary differential equation (ODE) with a vector field. We interpret an ordinary DNN as a transport map or an Euler broken line approximation of the flow. Technically speaking, a dynamical system is a natural model for the nested feature maps. In addition, it opens a new way to the coordinate-free treatment of DNNs by avoiding the redundant parametrization of DNNs. Following Wasserstein geometry, we analyze a flow in three aspects: dynamical system, continuity equation, and Wasserstein gradient flow. A key finding is that we specified a series of transport maps of the denoising autoencoder (DAE), which is a cornerstone for the development of deep learning. Starting from the shallow DAE, this paper develops three topics: the transport map of the deep DAE, the equivalence between the stacked DAE and the composition of DAEs, and the development of the double continuum limit or the integral representation of the flow representation. As partial answers to the research questions, we found that deeper DAEs converge faster and the extracted features are better; in addition, a deep Gaussian DAE transports mass to decrease the Shannon entropy of the data distribution. We expect that further investigations on these questions lead to the development of an interpretable and principled alternatives to DNNs." @default.
• W2899297079 created "2018-11-09" @default.
• W2899297079 creator A5036537988 @default.
• W2899297079 creator A5091474846 @default.
• W2899297079 date "2019-01-01" @default.
• W2899297079 modified "2023-09-23" @default.
• W2899297079 title "Transport analysis of infinitely deep neural network" @default.
• W2899297079 cites W1511694993 @default.
• W2899297079 cites W1567512734 @default.
• W2899297079 cites W1593038947 @default.
• W2899297079 cites W1836465849 @default.
• W2899297079 cites W1849277567 @default.
• W2899297079 cites W1899249567 @default.
• W2899297079 cites W1959608418 @default.
• W2899297079 cites W2013035813 @default.
• W2899297079 cites W2025768430 @default.
• W2899297079 cites W2032814004 @default.
• W2899297079 cites W2056860084 @default.
• W2899297079 cites W2101695397 @default.
• W2899297079 cites W2106751088 @default.
• W2899297079 cites W2110798204 @default.
• W2899297079 cites W2126398289 @default.
• W2899297079 cites W2129069237 @default.
• W2899297079 cites W2132883347 @default.
• W2899297079 cites W2145094598 @default.
• W2899297079 cites W2145287260 @default.
• W2899297079 cites W2146766088 @default.
• W2899297079 cites W2146989110 @default.
• W2899297079 cites W2159528849 @default.
• W2899297079 cites W2163922914 @default.
• W2899297079 cites W2166116275 @default.
• W2899297079 cites W2167967601 @default.
• W2899297079 cites W2194775991 @default.
• W2899297079 cites W2218318129 @default.
• W2899297079 cites W2281746805 @default.
• W2899297079 cites W2296319761 @default.
• W2899297079 cites W2399994860 @default.
• W2899297079 cites W2511837229 @default.
• W2899297079 cites W2528305538 @default.
• W2899297079 cites W2550848904 @default.
• W2899297079 cites W2566079294 @default.
• W2899297079 cites W2594458255 @default.
• W2899297079 cites W2606321545 @default.
• W2899297079 cites W2608609325 @default.
• W2899297079 cites W2614634292 @default.
• W2899297079 cites W2618530766 @default.
• W2899297079 cites W2739748921 @default.
• W2899297079 cites W2752860283 @default.
• W2899297079 cites W2771296525 @default.
• W2899297079 cites W2788800397 @default.
• W2899297079 cites W2789088344 @default.
• W2899297079 cites W2794272078 @default.
• W2899297079 cites W2798865883 @default.
• W2899297079 cites W2803636134 @default.
• W2899297079 cites W2804589149 @default.
• W2899297079 cites W2808746463 @default.
• W2899297079 cites W2895899761 @default.
• W2899297079 cites W2898001716 @default.
• W2899297079 cites W2952881492 @default.
• W2899297079 cites W2953256564 @default.
• W2899297079 cites W2953267151 @default.
• W2899297079 cites W2962742960 @default.
• W2899297079 cites W2962845550 @default.
• W2899297079 cites W2963302407 @default.
• W2899297079 cites W2963446085 @default.
• W2899297079 cites W2963534251 @default.
• W2899297079 cites W2963684275 @default.
• W2899297079 cites W2963709899 @default.
• W2899297079 cites W299440670 @default.
• W2899297079 cites W3022436500 @default.
• W2899297079 cites W3098011980 @default.
• W2899297079 cites W3101806332 @default.
• W2899297079 cites W3105432754 @default.
• W2899297079 cites W385466589 @default.
• W2899297079 cites W2180383608 @default.
• W2899297079 hasPublicationYear "2019" @default.
• W2899297079 type Work @default.
• W2899297079 sameAs 2899297079 @default.
• W2899297079 citedByCount "13" @default.
• W2899297079 countsByYear W28992970792019 @default.
• W2899297079 countsByYear W28992970792020 @default.
• W2899297079 countsByYear W28992970792021 @default.
• W2899297079 countsByYear W28992970792022 @default.
• W2899297079 crossrefType "journal-article" @default.
• W2899297079 hasAuthorship W2899297079A5036537988 @default.
• W2899297079 hasAuthorship W2899297079A5091474846 @default.
• W2899297079 hasConcept C11413529 @default.
• W2899297079 hasConcept C134306372 @default.
• W2899297079 hasConcept C167879884 @default.
• W2899297079 hasConcept C17744445 @default.
• W2899297079 hasConcept C199539241 @default.
• W2899297079 hasConcept C2524010 @default.
• W2899297079 hasConcept C2776359362 @default.
• W2899297079 hasConcept C28826006 @default.
• W2899297079 hasConcept C33923547 @default.
• W2899297079 hasConcept C34862557 @default.
• W2899297079 hasConcept C38349280 @default.
• W2899297079 hasConcept C41008148 @default.

• next