Matches in SemOpenAlex for { <https://semopenalex.org/work/W3016232661> ?p ?o ?g. }
- W3016232661 abstract "Differentiable Neural Architecture Search (DNAS) has demonstrated great success in designing state-of-the-art, efficient neural networks. However, DARTS-based DNAS's search space is small when compared to other search methods', since all candidate network layers must be explicitly instantiated in memory. To address this bottleneck, we propose a memory and computationally efficient DNAS variant: DMaskingNAS. This algorithm expands the search space by up to $10^{14}times$ over conventional DNAS, supporting searches over spatial and channel dimensions that are otherwise prohibitively expensive: input resolution and number of filters. We propose a masking mechanism for feature map reuse, so that memory and computational costs stay nearly constant as the search space expands. Furthermore, we employ effective shape propagation to maximize per-FLOP or per-parameter accuracy. The searched FBNetV2s yield state-of-the-art performance when compared with all previous architectures. With up to 421$times$ less search cost, DMaskingNAS finds models with 0.9% higher accuracy, 15% fewer FLOPs than MobileNetV3-Small; and with similar accuracy but 20% fewer FLOPs than Efficient-B0. Furthermore, our FBNetV2 outperforms MobileNetV3 by 2.6% in accuracy, with equivalent model size. FBNetV2 models are open-sourced at https://github.com/facebookresearch/mobile-vision." @default.
- W3016232661 created "2020-04-17" @default.
- W3016232661 creator A5001072947 @default.
- W3016232661 creator A5018305440 @default.
- W3016232661 creator A5024785349 @default.
- W3016232661 creator A5035953329 @default.
- W3016232661 creator A5036222171 @default.
- W3016232661 creator A5037373363 @default.
- W3016232661 creator A5048668303 @default.
- W3016232661 creator A5057613852 @default.
- W3016232661 creator A5072427753 @default.
- W3016232661 creator A5073512879 @default.
- W3016232661 creator A5080137972 @default.
- W3016232661 creator A5084821923 @default.
- W3016232661 date "2020-04-12" @default.
- W3016232661 modified "2023-10-18" @default.
- W3016232661 title "FBNetV2: Differentiable Neural Architecture Search for Spatial and Channel Dimensions" @default.
- W3016232661 cites W1522301498 @default.
- W3016232661 cites W2108598243 @default.
- W3016232661 cites W2194775991 @default.
- W3016232661 cites W2279098554 @default.
- W3016232661 cites W2547875792 @default.
- W3016232661 cites W2553303224 @default.
- W3016232661 cites W2594529350 @default.
- W3016232661 cites W2612445135 @default.
- W3016232661 cites W2736941579 @default.
- W3016232661 cites W2771727678 @default.
- W3016232661 cites W2783000019 @default.
- W3016232661 cites W2785366763 @default.
- W3016232661 cites W2885311373 @default.
- W3016232661 cites W2905672847 @default.
- W3016232661 cites W2932077855 @default.
- W3016232661 cites W2943899653 @default.
- W3016232661 cites W2944779197 @default.
- W3016232661 cites W2946948417 @default.
- W3016232661 cites W2950837708 @default.
- W3016232661 cites W2951104886 @default.
- W3016232661 cites W2951975363 @default.
- W3016232661 cites W2951977814 @default.
- W3016232661 cites W2962851801 @default.
- W3016232661 cites W2962861284 @default.
- W3016232661 cites W2963000224 @default.
- W3016232661 cites W2963087201 @default.
- W3016232661 cites W2963136578 @default.
- W3016232661 cites W2963420686 @default.
- W3016232661 cites W2963446712 @default.
- W3016232661 cites W2963674932 @default.
- W3016232661 cites W2963766446 @default.
- W3016232661 cites W2964259004 @default.
- W3016232661 cites W2964299589 @default.
- W3016232661 cites W2967733054 @default.
- W3016232661 cites W2980137827 @default.
- W3016232661 cites W2982479999 @default.
- W3016232661 cites W2997768846 @default.
- W3016232661 cites W3030728803 @default.
- W3016232661 doi "https://doi.org/10.48550/arxiv.2004.05565" @default.
- W3016232661 hasPublicationYear "2020" @default.
- W3016232661 type Work @default.
- W3016232661 sameAs 3016232661 @default.
- W3016232661 citedByCount "19" @default.
- W3016232661 countsByYear W30162326612019 @default.
- W3016232661 countsByYear W30162326612020 @default.
- W3016232661 countsByYear W30162326612021 @default.
- W3016232661 crossrefType "posted-content" @default.
- W3016232661 hasAuthorship W3016232661A5001072947 @default.
- W3016232661 hasAuthorship W3016232661A5018305440 @default.
- W3016232661 hasAuthorship W3016232661A5024785349 @default.
- W3016232661 hasAuthorship W3016232661A5035953329 @default.
- W3016232661 hasAuthorship W3016232661A5036222171 @default.
- W3016232661 hasAuthorship W3016232661A5037373363 @default.
- W3016232661 hasAuthorship W3016232661A5048668303 @default.
- W3016232661 hasAuthorship W3016232661A5057613852 @default.
- W3016232661 hasAuthorship W3016232661A5072427753 @default.
- W3016232661 hasAuthorship W3016232661A5073512879 @default.
- W3016232661 hasAuthorship W3016232661A5080137972 @default.
- W3016232661 hasAuthorship W3016232661A5084821923 @default.
- W3016232661 hasBestOaLocation W30162326611 @default.
- W3016232661 hasConcept C11413529 @default.
- W3016232661 hasConcept C125583679 @default.
- W3016232661 hasConcept C127162648 @default.
- W3016232661 hasConcept C134306372 @default.
- W3016232661 hasConcept C138885662 @default.
- W3016232661 hasConcept C142362112 @default.
- W3016232661 hasConcept C149635348 @default.
- W3016232661 hasConcept C153349607 @default.
- W3016232661 hasConcept C154945302 @default.
- W3016232661 hasConcept C162324750 @default.
- W3016232661 hasConcept C173608175 @default.
- W3016232661 hasConcept C175444787 @default.
- W3016232661 hasConcept C202615002 @default.
- W3016232661 hasConcept C21782646 @default.
- W3016232661 hasConcept C2776401178 @default.
- W3016232661 hasConcept C2777402240 @default.
- W3016232661 hasConcept C2780513914 @default.
- W3016232661 hasConcept C31258907 @default.
- W3016232661 hasConcept C33923547 @default.
- W3016232661 hasConcept C3826847 @default.
- W3016232661 hasConcept C41008148 @default.
- W3016232661 hasConcept C41895202 @default.
- W3016232661 hasConcept C50644808 @default.