Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385800833> ?p ?o ?g. }
Showing items 1 to 78 of
78
with 100 items per page.
- W4385800833 abstract "Brain-inspired event-driven processors execute deep neural networks (DNNs) in a sparsity-aware manner. Specifically, if more zeros are induced in the activation maps, less computation will be performed in the succeeding convolution layer. However, inducing activation sparsity in DNNs remains a challenge. To address this, we propose a training approach STAR (Sparse Thresholded Activation under partial-Regularization), which combines activation regularization with thresholding, to overcome the barrier of a single threshold- or regularization-based method in sparsity improvement. More precisely, we employ the sparse penalty on the near-zero activations to fit the activation learning behaviour in accuracy recovery, followed by thresholding to further suppress activations. Experimental results with SOTA networks (ResNet50/MobileNetV2, SSD, YOLOX and DeepLabV3+) on various datasets (Cifar-100, ImageNet, KITTI, VOC2007 and CityScapes) show that STAR can reduce on average 54% more activations compared to ReLU suppression. It outperforms the state-of-the-art by a significant margin of 35% in activation suppression without compromising accuracy loss. Additionally, a case study for a commercially-available event-driven hardware architecture, Neuronflow [29], demonstrates that the boosted activation sparsity in ResNet50 can be efficiently translated into latency reduction by up to 2.78×, FPS improvement by up to 2.80x, and energy savings by up to 2.09x. STAR elevates event-driven processors as a superior alternative to GPUs for Edge computing." @default.
- W4385800833 created "2023-08-15" @default.
- W4385800833 creator A5015357684 @default.
- W4385800833 creator A5063811722 @default.
- W4385800833 creator A5069322635 @default.
- W4385800833 creator A5085030425 @default.
- W4385800833 creator A5069685461 @default.
- W4385800833 date "2023-06-01" @default.
- W4385800833 modified "2023-10-16" @default.
- W4385800833 title "STAR: Sparse Thresholded Activation under partial-Regularization for Activation Sparsity Exploration" @default.
- W4385800833 cites W1604973310 @default.
- W4385800833 cites W2037227137 @default.
- W4385800833 cites W2108598243 @default.
- W4385800833 cites W2115579991 @default.
- W4385800833 cites W2194775991 @default.
- W4385800833 cites W2340897893 @default.
- W4385800833 cites W2516141709 @default.
- W4385800833 cites W2783525259 @default.
- W4385800833 cites W2883089954 @default.
- W4385800833 cites W2893813411 @default.
- W4385800833 cites W2964309882 @default.
- W4385800833 cites W2971115712 @default.
- W4385800833 cites W2988481835 @default.
- W4385800833 cites W3035770626 @default.
- W4385800833 cites W3139657805 @default.
- W4385800833 cites W4225301569 @default.
- W4385800833 cites W4236868170 @default.
- W4385800833 cites W4313484731 @default.
- W4385800833 doi "https://doi.org/10.1109/cvprw59228.2023.00479" @default.
- W4385800833 hasPublicationYear "2023" @default.
- W4385800833 type Work @default.
- W4385800833 citedByCount "0" @default.
- W4385800833 crossrefType "proceedings-article" @default.
- W4385800833 hasAuthorship W4385800833A5015357684 @default.
- W4385800833 hasAuthorship W4385800833A5063811722 @default.
- W4385800833 hasAuthorship W4385800833A5069322635 @default.
- W4385800833 hasAuthorship W4385800833A5069685461 @default.
- W4385800833 hasAuthorship W4385800833A5085030425 @default.
- W4385800833 hasConcept C11413529 @default.
- W4385800833 hasConcept C115961682 @default.
- W4385800833 hasConcept C153180895 @default.
- W4385800833 hasConcept C154945302 @default.
- W4385800833 hasConcept C191178318 @default.
- W4385800833 hasConcept C2776135515 @default.
- W4385800833 hasConcept C38365724 @default.
- W4385800833 hasConcept C41008148 @default.
- W4385800833 hasConcept C45347329 @default.
- W4385800833 hasConcept C50644808 @default.
- W4385800833 hasConcept C76155785 @default.
- W4385800833 hasConcept C82876162 @default.
- W4385800833 hasConceptScore W4385800833C11413529 @default.
- W4385800833 hasConceptScore W4385800833C115961682 @default.
- W4385800833 hasConceptScore W4385800833C153180895 @default.
- W4385800833 hasConceptScore W4385800833C154945302 @default.
- W4385800833 hasConceptScore W4385800833C191178318 @default.
- W4385800833 hasConceptScore W4385800833C2776135515 @default.
- W4385800833 hasConceptScore W4385800833C38365724 @default.
- W4385800833 hasConceptScore W4385800833C41008148 @default.
- W4385800833 hasConceptScore W4385800833C45347329 @default.
- W4385800833 hasConceptScore W4385800833C50644808 @default.
- W4385800833 hasConceptScore W4385800833C76155785 @default.
- W4385800833 hasConceptScore W4385800833C82876162 @default.
- W4385800833 hasLocation W43858008331 @default.
- W4385800833 hasOpenAccess W4385800833 @default.
- W4385800833 hasPrimaryLocation W43858008331 @default.
- W4385800833 hasRelatedWork W1622285688 @default.
- W4385800833 hasRelatedWork W2033000528 @default.
- W4385800833 hasRelatedWork W2295077625 @default.
- W4385800833 hasRelatedWork W2750730210 @default.
- W4385800833 hasRelatedWork W2770684271 @default.
- W4385800833 hasRelatedWork W2792520941 @default.
- W4385800833 hasRelatedWork W3103243916 @default.
- W4385800833 hasRelatedWork W4206745324 @default.
- W4385800833 hasRelatedWork W4299814392 @default.
- W4385800833 hasRelatedWork W2151415411 @default.
- W4385800833 isParatext "false" @default.
- W4385800833 isRetracted "false" @default.
- W4385800833 workType "article" @default.