Matches in SemOpenAlex for { <https://semopenalex.org/work/W3200312342> ?p ?o ?g. }
- W3200312342 abstract "Machine/deep-learning (ML/DL) based techniques are emerging as a driving force behind many cutting-edge technologies, achieving high accuracy on computer vision workloads such as image classification and object detection. However, training these models involving large parameters is both time-consuming and energy-hogging. In this regard, several prior works have advocated for sparsity to speed up the of DL training and more so, the inference phase. This work begins with the observation that during training, sparsity in the forward and backward passes are correlated. In that context, we investigate two types of sparsity (input and output type) inherent in gradient descent-based optimization algorithms and propose a hardware micro-architecture to leverage the same. Our experimental results use five state-of-the-art CNN models on the Imagenet dataset, and show back propagation speedups in the range of 1.69$times$ to 5.43$times$, compared to the dense baseline execution. By exploiting sparsity in both the forward and backward passes, speedup improvements range from 1.68$times$ to 3.30$times$ over the sparsity-agnostic baseline execution. Our work also achieves significant reduction in training iteration time over several previously proposed dense as well as sparse accelerator based platforms, in addition to achieving order of magnitude energy efficiency improvements over GPU based execution." @default.
- W3200312342 created "2021-09-27" @default.
- W3200312342 creator A5007116603 @default.
- W3200312342 creator A5054027488 @default.
- W3200312342 creator A5054146657 @default.
- W3200312342 creator A5058385222 @default.
- W3200312342 creator A5067815274 @default.
- W3200312342 creator A5071846463 @default.
- W3200312342 creator A5075012741 @default.
- W3200312342 creator A5076019669 @default.
- W3200312342 date "2021-09-16" @default.
- W3200312342 modified "2023-10-16" @default.
- W3200312342 title "Exploiting Activation based Gradient Output Sparsity to Accelerate Backpropagation in CNNs." @default.
- W3200312342 cites W1686810756 @default.
- W3200312342 cites W1903029394 @default.
- W3200312342 cites W1987840949 @default.
- W3200312342 cites W2000967104 @default.
- W3200312342 cites W2038071684 @default.
- W3200312342 cites W2048266589 @default.
- W3200312342 cites W2067523571 @default.
- W3200312342 cites W2097117768 @default.
- W3200312342 cites W2117539524 @default.
- W3200312342 cites W2124007994 @default.
- W3200312342 cites W2141125852 @default.
- W3200312342 cites W2152839228 @default.
- W3200312342 cites W2156387975 @default.
- W3200312342 cites W2285660444 @default.
- W3200312342 cites W2289252105 @default.
- W3200312342 cites W2302255633 @default.
- W3200312342 cites W2419448466 @default.
- W3200312342 cites W2510439625 @default.
- W3200312342 cites W2516141709 @default.
- W3200312342 cites W2565851976 @default.
- W3200312342 cites W2605019159 @default.
- W3200312342 cites W2606722458 @default.
- W3200312342 cites W2612445135 @default.
- W3200312342 cites W2613718673 @default.
- W3200312342 cites W2625457103 @default.
- W3200312342 cites W2767286248 @default.
- W3200312342 cites W2787752464 @default.
- W3200312342 cites W2798724095 @default.
- W3200312342 cites W2883542588 @default.
- W3200312342 cites W2944850220 @default.
- W3200312342 cites W2955425717 @default.
- W3200312342 cites W2963000224 @default.
- W3200312342 cites W2963446712 @default.
- W3200312342 cites W2963674932 @default.
- W3200312342 cites W2964080840 @default.
- W3200312342 cites W2964261616 @default.
- W3200312342 cites W2979439447 @default.
- W3200312342 cites W2990510620 @default.
- W3200312342 cites W3047872496 @default.
- W3200312342 cites W3092357178 @default.
- W3200312342 hasPublicationYear "2021" @default.
- W3200312342 type Work @default.
- W3200312342 sameAs 3200312342 @default.
- W3200312342 citedByCount "0" @default.
- W3200312342 crossrefType "posted-content" @default.
- W3200312342 hasAuthorship W3200312342A5007116603 @default.
- W3200312342 hasAuthorship W3200312342A5054027488 @default.
- W3200312342 hasAuthorship W3200312342A5054146657 @default.
- W3200312342 hasAuthorship W3200312342A5058385222 @default.
- W3200312342 hasAuthorship W3200312342A5067815274 @default.
- W3200312342 hasAuthorship W3200312342A5071846463 @default.
- W3200312342 hasAuthorship W3200312342A5075012741 @default.
- W3200312342 hasAuthorship W3200312342A5076019669 @default.
- W3200312342 hasConcept C108583219 @default.
- W3200312342 hasConcept C111335779 @default.
- W3200312342 hasConcept C111368507 @default.
- W3200312342 hasConcept C113775141 @default.
- W3200312342 hasConcept C11413529 @default.
- W3200312342 hasConcept C12725497 @default.
- W3200312342 hasConcept C127313418 @default.
- W3200312342 hasConcept C151730666 @default.
- W3200312342 hasConcept C153083717 @default.
- W3200312342 hasConcept C153258448 @default.
- W3200312342 hasConcept C154945302 @default.
- W3200312342 hasConcept C155032097 @default.
- W3200312342 hasConcept C159985019 @default.
- W3200312342 hasConcept C173608175 @default.
- W3200312342 hasConcept C192562407 @default.
- W3200312342 hasConcept C204323151 @default.
- W3200312342 hasConcept C2524010 @default.
- W3200312342 hasConcept C2776214188 @default.
- W3200312342 hasConcept C2779343474 @default.
- W3200312342 hasConcept C33923547 @default.
- W3200312342 hasConcept C41008148 @default.
- W3200312342 hasConcept C50644808 @default.
- W3200312342 hasConcept C68339613 @default.
- W3200312342 hasConcept C86803240 @default.
- W3200312342 hasConceptScore W3200312342C108583219 @default.
- W3200312342 hasConceptScore W3200312342C111335779 @default.
- W3200312342 hasConceptScore W3200312342C111368507 @default.
- W3200312342 hasConceptScore W3200312342C113775141 @default.
- W3200312342 hasConceptScore W3200312342C11413529 @default.
- W3200312342 hasConceptScore W3200312342C12725497 @default.
- W3200312342 hasConceptScore W3200312342C127313418 @default.
- W3200312342 hasConceptScore W3200312342C151730666 @default.
- W3200312342 hasConceptScore W3200312342C153083717 @default.
- W3200312342 hasConceptScore W3200312342C153258448 @default.