Matches in SemOpenAlex for { <https://semopenalex.org/work/W2953350231> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W2953350231 abstract "It is well known that Convolutional Neural Networks (CNNs) have significant redundancy in their filter weights. Various methods have been proposed in the literature to compress trained CNNs. These include techniques like pruning weights, filter quantization and representing filters in terms of a basis functions. Our approach falls in this latter class of strategies, but is distinct in that that we show both compressed learning and representation can be achieved without significant modifications of popular CNN architectures. Specifically, any convolution layer of the CNN is easily replaced by two successive convolution layers: the first is a set of fixed filters (that represent the knowledge space of the entire layer and do not change), which is followed by a layer of one-dimensional filters (that represent the learned knowledge in this space). For the pre-trained networks, the fixed layer is just the truncated eigen-decompositions of the original filters. The 1D filters are initialized as the weights of linear combination, but are fine-tuned to recover any performance loss due to the truncation. For training networks from scratch, we use a set of random orthogonal fixed filters (that never change), and learn the 1D weight vector directly from the labeled data. Our method substantially reduces i) the number of learnable parameters during training, and ii) the number of multiplication operations and filter storage requirements during implementation. It does so without requiring any special operators in the convolution layer, and extends to all known popular CNN architectures. We apply our method to four well known network architectures trained with three different data sets. Results show a consistent reduction in i) the number of operations by up to a factor of 5, and ii) number of learnable parameters by up to a factor of 18, with less than 3% drop in performance on the CIFAR100 dataset." @default.
- W2953350231 created "2019-06-27" @default.
- W2953350231 creator A5055793435 @default.
- W2953350231 creator A5090139074 @default.
- W2953350231 date "2019-06-11" @default.
- W2953350231 modified "2023-09-24" @default.
- W2953350231 title "BasisConv: A method for compressed representation and learning in CNNs" @default.
- W2953350231 cites W1724438581 @default.
- W2953350231 cites W1821462560 @default.
- W2953350231 cites W2114766824 @default.
- W2953350231 cites W2125389748 @default.
- W2953350231 cites W2134797427 @default.
- W2953350231 cites W2167215970 @default.
- W2953350231 cites W2172166488 @default.
- W2953350231 cites W2194775991 @default.
- W2953350231 cites W2233116163 @default.
- W2953350231 cites W2260663238 @default.
- W2953350231 cites W2300242332 @default.
- W2953350231 cites W2618530766 @default.
- W2953350231 cites W2766839578 @default.
- W2953350231 cites W2950967261 @default.
- W2953350231 cites W2962835968 @default.
- W2953350231 cites W2963076838 @default.
- W2953350231 cites W2963114950 @default.
- W2953350231 cites W2963341152 @default.
- W2953350231 cites W2963446712 @default.
- W2953350231 cites W2963674932 @default.
- W2953350231 cites W2964217848 @default.
- W2953350231 cites W2964299589 @default.
- W2953350231 cites W587794757 @default.
- W2953350231 hasPublicationYear "2019" @default.
- W2953350231 type Work @default.
- W2953350231 sameAs 2953350231 @default.
- W2953350231 citedByCount "1" @default.
- W2953350231 countsByYear W29533502312020 @default.
- W2953350231 crossrefType "posted-content" @default.
- W2953350231 hasAuthorship W2953350231A5055793435 @default.
- W2953350231 hasAuthorship W2953350231A5090139074 @default.
- W2953350231 hasConcept C106131492 @default.
- W2953350231 hasConcept C111919701 @default.
- W2953350231 hasConcept C11413529 @default.
- W2953350231 hasConcept C152124472 @default.
- W2953350231 hasConcept C153180895 @default.
- W2953350231 hasConcept C154945302 @default.
- W2953350231 hasConcept C31972630 @default.
- W2953350231 hasConcept C33923547 @default.
- W2953350231 hasConcept C41008148 @default.
- W2953350231 hasConcept C45347329 @default.
- W2953350231 hasConcept C50644808 @default.
- W2953350231 hasConcept C81363708 @default.
- W2953350231 hasConceptScore W2953350231C106131492 @default.
- W2953350231 hasConceptScore W2953350231C111919701 @default.
- W2953350231 hasConceptScore W2953350231C11413529 @default.
- W2953350231 hasConceptScore W2953350231C152124472 @default.
- W2953350231 hasConceptScore W2953350231C153180895 @default.
- W2953350231 hasConceptScore W2953350231C154945302 @default.
- W2953350231 hasConceptScore W2953350231C31972630 @default.
- W2953350231 hasConceptScore W2953350231C33923547 @default.
- W2953350231 hasConceptScore W2953350231C41008148 @default.
- W2953350231 hasConceptScore W2953350231C45347329 @default.
- W2953350231 hasConceptScore W2953350231C50644808 @default.
- W2953350231 hasConceptScore W2953350231C81363708 @default.
- W2953350231 hasLocation W29533502311 @default.
- W2953350231 hasOpenAccess W2953350231 @default.
- W2953350231 hasPrimaryLocation W29533502311 @default.
- W2953350231 hasRelatedWork W2554242204 @default.
- W2953350231 hasRelatedWork W2593245696 @default.
- W2953350231 hasRelatedWork W2735189290 @default.
- W2953350231 hasRelatedWork W2737371731 @default.
- W2953350231 hasRelatedWork W2805022637 @default.
- W2953350231 hasRelatedWork W2808145095 @default.
- W2953350231 hasRelatedWork W2808868252 @default.
- W2953350231 hasRelatedWork W2945335799 @default.
- W2953350231 hasRelatedWork W2951283089 @default.
- W2953350231 hasRelatedWork W2970714388 @default.
- W2953350231 hasRelatedWork W2979596232 @default.
- W2953350231 hasRelatedWork W2980807300 @default.
- W2953350231 hasRelatedWork W2988887648 @default.
- W2953350231 hasRelatedWork W2989991345 @default.
- W2953350231 hasRelatedWork W3002347511 @default.
- W2953350231 hasRelatedWork W3112671111 @default.
- W2953350231 hasRelatedWork W3124157622 @default.
- W2953350231 hasRelatedWork W3168810591 @default.
- W2953350231 hasRelatedWork W3174055538 @default.
- W2953350231 hasRelatedWork W3116076050 @default.
- W2953350231 isParatext "false" @default.
- W2953350231 isRetracted "false" @default.
- W2953350231 magId "2953350231" @default.
- W2953350231 workType "article" @default.