Matches in SemOpenAlex for { <https://semopenalex.org/work/W3167838442> ?p ?o ?g. }
Showing items 1 to 100 of
100
with 100 items per page.
- W3167838442 abstract "The Neural Tangent Kernel (NTK) characterizes the behavior of infinitely-wide neural networks trained under least squares loss by gradient descent. Recent works also report that NTK regression can outperform finitely-wide neural networks trained on small-scale datasets. However, the computational complexity of kernel methods has limited its use in large-scale learning tasks. To accelerate learning with NTK, we design a near input-sparsity time approximation algorithm for NTK, by sketching the polynomial expansions of arc-cosine kernels: our sketch for the convolutional counterpart of NTK (CNTK) can transform any image using a linear runtime in the number of pixels. Furthermore, we prove a spectral approximation guarantee for the NTK matrix, by combining random features (based on leverage score sampling) of the arc-cosine kernels with a sketching algorithm. We benchmark our methods on various large-scale regression and classification tasks and show that a linear regressor trained on our CNTK features matches the accuracy of exact CNTK on CIFAR-10 dataset while achieving 150x speedup." @default.
- W3167838442 created "2021-06-22" @default.
- W3167838442 creator A5014238886 @default.
- W3167838442 creator A5022094360 @default.
- W3167838442 creator A5035397475 @default.
- W3167838442 creator A5062405513 @default.
- W3167838442 creator A5072151198 @default.
- W3167838442 creator A5082060594 @default.
- W3167838442 date "2021-06-15" @default.
- W3167838442 modified "2023-09-28" @default.
- W3167838442 title "Scaling Neural Tangent Kernels via Sketching and Random Features" @default.
- W3167838442 cites W2093813380 @default.
- W3167838442 cites W2121689290 @default.
- W3167838442 cites W2144902422 @default.
- W3167838442 cites W2151683004 @default.
- W3167838442 cites W2157988812 @default.
- W3167838442 cites W2167608136 @default.
- W3167838442 cites W2182347905 @default.
- W3167838442 cites W2809090039 @default.
- W3167838442 cites W2900959181 @default.
- W3167838442 cites W2910655610 @default.
- W3167838442 cites W2913473169 @default.
- W3167838442 cites W2952204734 @default.
- W3167838442 cites W2962685794 @default.
- W3167838442 cites W2962939986 @default.
- W3167838442 cites W2963013450 @default.
- W3167838442 cites W2963095610 @default.
- W3167838442 cites W2963239103 @default.
- W3167838442 cites W2963323437 @default.
- W3167838442 cites W2963459001 @default.
- W3167838442 cites W2963624229 @default.
- W3167838442 cites W2963695615 @default.
- W3167838442 cites W2963923734 @default.
- W3167838442 cites W2964052793 @default.
- W3167838442 cites W2964098911 @default.
- W3167838442 cites W2964161337 @default.
- W3167838442 cites W2970217468 @default.
- W3167838442 cites W2970239942 @default.
- W3167838442 cites W2970332347 @default.
- W3167838442 cites W2971043187 @default.
- W3167838442 cites W2994872659 @default.
- W3167838442 cites W3001361757 @default.
- W3167838442 cites W3034322278 @default.
- W3167838442 cites W3034510050 @default.
- W3167838442 cites W3034761945 @default.
- W3167838442 cites W3034979923 @default.
- W3167838442 cites W3089200980 @default.
- W3167838442 cites W3098047114 @default.
- W3167838442 cites W3099842580 @default.
- W3167838442 cites W3100387437 @default.
- W3167838442 cites W3118608800 @default.
- W3167838442 cites W3120740533 @default.
- W3167838442 hasPublicationYear "2021" @default.
- W3167838442 type Work @default.
- W3167838442 sameAs 3167838442 @default.
- W3167838442 citedByCount "0" @default.
- W3167838442 crossrefType "posted-content" @default.
- W3167838442 hasAuthorship W3167838442A5014238886 @default.
- W3167838442 hasAuthorship W3167838442A5022094360 @default.
- W3167838442 hasAuthorship W3167838442A5035397475 @default.
- W3167838442 hasAuthorship W3167838442A5062405513 @default.
- W3167838442 hasAuthorship W3167838442A5072151198 @default.
- W3167838442 hasAuthorship W3167838442A5082060594 @default.
- W3167838442 hasConcept C11413529 @default.
- W3167838442 hasConcept C118615104 @default.
- W3167838442 hasConcept C138187205 @default.
- W3167838442 hasConcept C153083717 @default.
- W3167838442 hasConcept C154945302 @default.
- W3167838442 hasConcept C2524010 @default.
- W3167838442 hasConcept C33923547 @default.
- W3167838442 hasConcept C41008148 @default.
- W3167838442 hasConcept C74193536 @default.
- W3167838442 hasConcept C81363708 @default.
- W3167838442 hasConcept C99844830 @default.
- W3167838442 hasConceptScore W3167838442C11413529 @default.
- W3167838442 hasConceptScore W3167838442C118615104 @default.
- W3167838442 hasConceptScore W3167838442C138187205 @default.
- W3167838442 hasConceptScore W3167838442C153083717 @default.
- W3167838442 hasConceptScore W3167838442C154945302 @default.
- W3167838442 hasConceptScore W3167838442C2524010 @default.
- W3167838442 hasConceptScore W3167838442C33923547 @default.
- W3167838442 hasConceptScore W3167838442C41008148 @default.
- W3167838442 hasConceptScore W3167838442C74193536 @default.
- W3167838442 hasConceptScore W3167838442C81363708 @default.
- W3167838442 hasConceptScore W3167838442C99844830 @default.
- W3167838442 hasOpenAccess W3167838442 @default.
- W3167838442 hasRelatedWork W1167735 @default.
- W3167838442 hasRelatedWork W12139544 @default.
- W3167838442 hasRelatedWork W12560482 @default.
- W3167838442 hasRelatedWork W2094762 @default.
- W3167838442 hasRelatedWork W3860106 @default.
- W3167838442 hasRelatedWork W5924924 @default.
- W3167838442 hasRelatedWork W6717794 @default.
- W3167838442 hasRelatedWork W7867045 @default.
- W3167838442 hasRelatedWork W9190101 @default.
- W3167838442 hasRelatedWork W9877987 @default.
- W3167838442 isParatext "false" @default.
- W3167838442 isRetracted "false" @default.
- W3167838442 magId "3167838442" @default.
- W3167838442 workType "article" @default.