Matches in SemOpenAlex for { <https://semopenalex.org/work/W3134864592> ?p ?o ?g. }
Showing items 1 to 84 of
84
with 100 items per page.
- W3134864592 endingPage "240" @default.
- W3134864592 startingPage "229" @default.
- W3134864592 abstract "Systolic arrays are a promising computing concept which is in particular inline with CMOS technology trends and linear algebra operations found in the processing of artificial neural networks. The recent success of such deep learning methods in a wide set of applications has led to a variety of models, which albeit conceptual similar as based on convolutions and fully-connected layers, in detail show a huge diversity in operations due to a large design space: An operand’s dimension varies substantially since it depends on design principles such as receptive field size, number of features, striding, dilating and grouping of features. Last, recent networks extent previously plain feedforward models by various connectivity, such as in ResNet or DenseNet. The problem of choosing an optimal systolic array configuration cannot be solved analytically, thus instead methods and tools are required that facilitate a fast and accurate reasoning about optimality in terms of total cycles, utilization, and amount of data movements. In this work we introduce Camuy, a lightweight model of a weight-stationary systolic array for linear algebra operations that allows quick explorations of different configurations, such as systolic array dimensions and input/output bitwidths. Camuy aids accelerator designers in either finding optimal configurations for a particular network architecture or for robust performance across a variety of network architectures. It offers simple integration into existing machine learning tool stacks (e.g TensorFlow) through custom operators. We present an analysis of popular DNN models to illustrate how it can estimate required cycles, data movement costs, as well as systolic array utilization, and show how the progress in network architecture design impacts the efficiency of inference on accelerators based on systolic arrays." @default.
- W3134864592 created "2021-03-15" @default.
- W3134864592 creator A5014251134 @default.
- W3134864592 creator A5056625768 @default.
- W3134864592 creator A5073596132 @default.
- W3134864592 date "2020-01-01" @default.
- W3134864592 modified "2023-10-16" @default.
- W3134864592 title "On the Difficulty of Designing Processor Arrays for Deep Neural Networks" @default.
- W3134864592 cites W1999085092 @default.
- W3134864592 cites W2017369466 @default.
- W3134864592 cites W2126105956 @default.
- W3134864592 cites W2194775991 @default.
- W3134864592 cites W2442974303 @default.
- W3134864592 cites W2549139847 @default.
- W3134864592 cites W2963446712 @default.
- W3134864592 cites W3007158167 @default.
- W3134864592 cites W3036079062 @default.
- W3134864592 cites W4231250608 @default.
- W3134864592 cites W4240916550 @default.
- W3134864592 cites W4288083528 @default.
- W3134864592 doi "https://doi.org/10.1007/978-3-030-66770-2_17" @default.
- W3134864592 hasPublicationYear "2020" @default.
- W3134864592 type Work @default.
- W3134864592 sameAs 3134864592 @default.
- W3134864592 citedByCount "0" @default.
- W3134864592 crossrefType "book-chapter" @default.
- W3134864592 hasAuthorship W3134864592A5014251134 @default.
- W3134864592 hasAuthorship W3134864592A5056625768 @default.
- W3134864592 hasAuthorship W3134864592A5073596132 @default.
- W3134864592 hasBestOaLocation W31348645922 @default.
- W3134864592 hasConcept C108583219 @default.
- W3134864592 hasConcept C111472728 @default.
- W3134864592 hasConcept C113775141 @default.
- W3134864592 hasConcept C136197465 @default.
- W3134864592 hasConcept C138885662 @default.
- W3134864592 hasConcept C14580979 @default.
- W3134864592 hasConcept C149635348 @default.
- W3134864592 hasConcept C150741067 @default.
- W3134864592 hasConcept C154945302 @default.
- W3134864592 hasConcept C202444582 @default.
- W3134864592 hasConcept C2780586882 @default.
- W3134864592 hasConcept C33676613 @default.
- W3134864592 hasConcept C33923547 @default.
- W3134864592 hasConcept C41008148 @default.
- W3134864592 hasConcept C50644808 @default.
- W3134864592 hasConcept C55526617 @default.
- W3134864592 hasConcept C9390403 @default.
- W3134864592 hasConceptScore W3134864592C108583219 @default.
- W3134864592 hasConceptScore W3134864592C111472728 @default.
- W3134864592 hasConceptScore W3134864592C113775141 @default.
- W3134864592 hasConceptScore W3134864592C136197465 @default.
- W3134864592 hasConceptScore W3134864592C138885662 @default.
- W3134864592 hasConceptScore W3134864592C14580979 @default.
- W3134864592 hasConceptScore W3134864592C149635348 @default.
- W3134864592 hasConceptScore W3134864592C150741067 @default.
- W3134864592 hasConceptScore W3134864592C154945302 @default.
- W3134864592 hasConceptScore W3134864592C202444582 @default.
- W3134864592 hasConceptScore W3134864592C2780586882 @default.
- W3134864592 hasConceptScore W3134864592C33676613 @default.
- W3134864592 hasConceptScore W3134864592C33923547 @default.
- W3134864592 hasConceptScore W3134864592C41008148 @default.
- W3134864592 hasConceptScore W3134864592C50644808 @default.
- W3134864592 hasConceptScore W3134864592C55526617 @default.
- W3134864592 hasConceptScore W3134864592C9390403 @default.
- W3134864592 hasLocation W31348645921 @default.
- W3134864592 hasLocation W31348645922 @default.
- W3134864592 hasOpenAccess W3134864592 @default.
- W3134864592 hasPrimaryLocation W31348645921 @default.
- W3134864592 hasRelatedWork W1585158103 @default.
- W3134864592 hasRelatedWork W1647292680 @default.
- W3134864592 hasRelatedWork W2055541046 @default.
- W3134864592 hasRelatedWork W2138767404 @default.
- W3134864592 hasRelatedWork W2731899572 @default.
- W3134864592 hasRelatedWork W2948510819 @default.
- W3134864592 hasRelatedWork W3037777028 @default.
- W3134864592 hasRelatedWork W4210298930 @default.
- W3134864592 hasRelatedWork W4224093850 @default.
- W3134864592 hasRelatedWork W4287752851 @default.
- W3134864592 isParatext "false" @default.
- W3134864592 isRetracted "false" @default.
- W3134864592 magId "3134864592" @default.
- W3134864592 workType "book-chapter" @default.