FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2785961331> ?p ?o ?g. }

• W2785961331 abstract "The state-of-the-art (SOTA) for mixed precision training is dominated by variants of low precision floating point operations, and in particular, FP16 accumulating into FP32 Micikevicius et al. (2017). On the other hand, while a lot of research has also happened in the domain of low and mixed-precision Integer training, these works either present results for non-SOTA networks (for instance only AlexNet for ImageNet-1K), or relatively small datasets (like CIFAR-10). In this work, we train state-of-the-art visual understanding neural networks on the ImageNet-1K dataset, with Integer operations on General Purpose (GP) hardware. In particular, we focus on Integer Fused-Multiply-and-Accumulate (FMA) operations which take two pairs of INT16 operands and accumulate results into an INT32 output.We propose a shared exponent representation of tensors and develop a Dynamic Fixed Point (DFP) scheme suitable for common neural network operations. The nuances of developing an efficient integer convolution kernel is examined, including methods to handle overflow of the INT32 accumulator. We implement CNN training for ResNet-50, GoogLeNet-v1, VGG-16 and AlexNet; and these networks achieve or exceed SOTA accuracy within the same number of iterations as their FP32 counterparts without any change in hyper-parameters and with a 1.8X improvement in end-to-end training throughput. To the best of our knowledge these results represent the first INT16 training results on GP hardware for ImageNet-1K dataset using SOTA CNNs and achieve highest reported accuracy using half-precision" @default.
• W2785961331 created "2018-02-23" @default.
• W2785961331 creator A5009364875 @default.
• W2785961331 creator A5018944355 @default.
• W2785961331 creator A5021152315 @default.
• W2785961331 creator A5022344235 @default.
• W2785961331 creator A5023420478 @default.
• W2785961331 creator A5023588595 @default.
• W2785961331 creator A5024066643 @default.
• W2785961331 creator A5024617712 @default.
• W2785961331 creator A5032238070 @default.
• W2785961331 creator A5034003919 @default.
• W2785961331 creator A5044770886 @default.
• W2785961331 creator A5052810907 @default.
• W2785961331 creator A5054664575 @default.
• W2785961331 creator A5057664789 @default.
• W2785961331 creator A5066366329 @default.
• W2785961331 creator A5068139008 @default.
• W2785961331 creator A5083700279 @default.
• W2785961331 date "2018-02-03" @default.
• W2785961331 modified "2023-09-29" @default.
• W2785961331 title "Mixed Precision Training of Convolutional Neural Networks using Integer Operations" @default.
• W2785961331 cites W1594170634 @default.
• W2785961331 cites W1686810756 @default.
• W2785961331 cites W1902934009 @default.
• W2785961331 cites W2108598243 @default.
• W2785961331 cites W2117539524 @default.
• W2785961331 cites W2177838837 @default.
• W2785961331 cites W2194775991 @default.
• W2785961331 cites W2763421725 @default.
• W2785961331 cites W2950094539 @default.
• W2785961331 cites W2950894517 @default.
• W2785961331 cites W2962786581 @default.
• W2785961331 hasPublicationYear "2018" @default.
• W2785961331 type Work @default.
• W2785961331 sameAs 2785961331 @default.
• W2785961331 citedByCount "31" @default.
• W2785961331 countsByYear W27859613312018 @default.
• W2785961331 countsByYear W27859613312019 @default.
• W2785961331 countsByYear W27859613312020 @default.
• W2785961331 countsByYear W27859613312021 @default.
• W2785961331 crossrefType "posted-content" @default.
• W2785961331 hasAuthorship W2785961331A5009364875 @default.
• W2785961331 hasAuthorship W2785961331A5018944355 @default.
• W2785961331 hasAuthorship W2785961331A5021152315 @default.
• W2785961331 hasAuthorship W2785961331A5022344235 @default.
• W2785961331 hasAuthorship W2785961331A5023420478 @default.
• W2785961331 hasAuthorship W2785961331A5023588595 @default.
• W2785961331 hasAuthorship W2785961331A5024066643 @default.
• W2785961331 hasAuthorship W2785961331A5024617712 @default.
• W2785961331 hasAuthorship W2785961331A5032238070 @default.
• W2785961331 hasAuthorship W2785961331A5034003919 @default.
• W2785961331 hasAuthorship W2785961331A5044770886 @default.
• W2785961331 hasAuthorship W2785961331A5052810907 @default.
• W2785961331 hasAuthorship W2785961331A5054664575 @default.
• W2785961331 hasAuthorship W2785961331A5057664789 @default.
• W2785961331 hasAuthorship W2785961331A5066366329 @default.
• W2785961331 hasAuthorship W2785961331A5068139008 @default.
• W2785961331 hasAuthorship W2785961331A5083700279 @default.
• W2785961331 hasConcept C11413529 @default.
• W2785961331 hasConcept C114614502 @default.
• W2785961331 hasConcept C119857082 @default.
• W2785961331 hasConcept C134306372 @default.
• W2785961331 hasConcept C153180895 @default.
• W2785961331 hasConcept C154945302 @default.
• W2785961331 hasConcept C199360897 @default.
• W2785961331 hasConcept C2984842247 @default.
• W2785961331 hasConcept C33923547 @default.
• W2785961331 hasConcept C41008148 @default.
• W2785961331 hasConcept C45347329 @default.
• W2785961331 hasConcept C50644808 @default.
• W2785961331 hasConcept C55526617 @default.
• W2785961331 hasConcept C74193536 @default.
• W2785961331 hasConcept C77618280 @default.
• W2785961331 hasConcept C81363708 @default.
• W2785961331 hasConcept C84211073 @default.
• W2785961331 hasConcept C9390403 @default.
• W2785961331 hasConcept C97137487 @default.
• W2785961331 hasConceptScore W2785961331C11413529 @default.
• W2785961331 hasConceptScore W2785961331C114614502 @default.
• W2785961331 hasConceptScore W2785961331C119857082 @default.
• W2785961331 hasConceptScore W2785961331C134306372 @default.
• W2785961331 hasConceptScore W2785961331C153180895 @default.
• W2785961331 hasConceptScore W2785961331C154945302 @default.
• W2785961331 hasConceptScore W2785961331C199360897 @default.
• W2785961331 hasConceptScore W2785961331C2984842247 @default.
• W2785961331 hasConceptScore W2785961331C33923547 @default.
• W2785961331 hasConceptScore W2785961331C41008148 @default.
• W2785961331 hasConceptScore W2785961331C45347329 @default.
• W2785961331 hasConceptScore W2785961331C50644808 @default.
• W2785961331 hasConceptScore W2785961331C55526617 @default.
• W2785961331 hasConceptScore W2785961331C74193536 @default.
• W2785961331 hasConceptScore W2785961331C77618280 @default.
• W2785961331 hasConceptScore W2785961331C81363708 @default.
• W2785961331 hasConceptScore W2785961331C84211073 @default.
• W2785961331 hasConceptScore W2785961331C9390403 @default.
• W2785961331 hasConceptScore W2785961331C97137487 @default.
• W2785961331 hasLocation W27859613311 @default.
• W2785961331 hasOpenAccess W2785961331 @default.
• W2785961331 hasPrimaryLocation W27859613311 @default.

• next