FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2553100439> ?p ?o ?g. }

• W2553100439 abstract "Depth is one of the key factors behind the success of convolutional neural networks (CNNs). Since ResNet, we are able to train very deep CNNs as the gradient vanishing issue has been largely addressed by the introduction of skip connections. However, we observe that, when the depth is very large, the intermediate layers (especially shallow layers) may fail to receive sufficient supervision from the loss due to the severe transformation through a long backpropagation path. As a result, the representation power of intermediate layers can be very weak and the model becomes very redundant with limited performance. In this paper, we first investigate the supervision vanishing issue in existing backpropagation (BP) methods. And then, we propose to address it via an effective method, called Multi-way BP (MW-BP), which relies on multiple auxiliary losses added to the intermediate layers of the network. The proposed MW-BP method can be applied to most deep architectures with slight modifications, such as ResNet and MobileNet. Our method often gives rise to much more compact models (denoted by Mw+Architecture) than existing methods. For example, MwResNet-44 with 44 layers performs better than ResNet-110 with 110 layers on CIFAR-10 and CIFAR-100. More critically, the resultant models even outperform the light models obtained by state-of-the-art model compression methods. Last, our method inherently produces multiple compact models with different depths at the same time, which is helpful for model selection." @default.
• W2553100439 created "2016-11-30" @default.
• W2553100439 creator A5003929021 @default.
• W2553100439 creator A5028024287 @default.
• W2553100439 creator A5032352025 @default.
• W2553100439 creator A5033588812 @default.
• W2553100439 creator A5058406424 @default.
• W2553100439 creator A5090234226 @default.
• W2553100439 date "2016-11-06" @default.
• W2553100439 modified "2023-09-27" @default.
• W2553100439 title "The Shallow End: Empowering Shallower Deep-Convolutional Networks through Auxiliary Outputs." @default.
• W2553100439 cites W1026270304 @default.
• W2553100439 cites W1567302070 @default.
• W2553100439 cites W1594587862 @default.
• W2553100439 cites W1665214252 @default.
• W2553100439 cites W1677182931 @default.
• W2553100439 cites W1690739335 @default.
• W2553100439 cites W1782590233 @default.
• W2553100439 cites W1849277567 @default.
• W2553100439 cites W1903029394 @default.
• W2553100439 cites W1950843348 @default.
• W2553100439 cites W2097117768 @default.
• W2553100439 cites W2112594540 @default.
• W2553100439 cites W2117130368 @default.
• W2553100439 cites W2117539524 @default.
• W2553100439 cites W2125556102 @default.
• W2553100439 cites W2133319764 @default.
• W2553100439 cites W2134797427 @default.
• W2553100439 cites W2140609507 @default.
• W2553100439 cites W2147800946 @default.
• W2553100439 cites W2154987621 @default.
• W2553100439 cites W2163605009 @default.
• W2553100439 cites W2194775991 @default.
• W2553100439 cites W2302255633 @default.
• W2553100439 cites W2331143823 @default.
• W2553100439 cites W2335728318 @default.
• W2553100439 cites W2339172597 @default.
• W2553100439 cites W2339890635 @default.
• W2553100439 cites W2395611524 @default.
• W2553100439 cites W2404498690 @default.
• W2553100439 cites W2408074187 @default.
• W2553100439 cites W2412782625 @default.
• W2553100439 cites W2515770085 @default.
• W2553100439 cites W2549139847 @default.
• W2553100439 cites W2612445135 @default.
• W2553100439 cites W2663800299 @default.
• W2553100439 cites W2919115771 @default.
• W2553100439 cites W2951603627 @default.
• W2553100439 cites W2962677625 @default.
• W2553100439 cites W2962835968 @default.
• W2553100439 cites W2962845550 @default.
• W2553100439 cites W2962965870 @default.
• W2553100439 cites W2963125010 @default.
• W2553100439 cites W2963145730 @default.
• W2553100439 cites W2963363373 @default.
• W2553100439 cites W2963377935 @default.
• W2553100439 cites W2963446712 @default.
• W2553100439 cites W2963466847 @default.
• W2553100439 cites W2963606038 @default.
• W2553100439 cites W2963671154 @default.
• W2553100439 cites W2963881378 @default.
• W2553100439 cites W2964137095 @default.
• W2553100439 cites W2964233199 @default.
• W2553100439 cites W2964303913 @default.
• W2553100439 cites W2964350391 @default.
• W2553100439 cites W3038058348 @default.
• W2553100439 cites W3118608800 @default.
• W2553100439 cites W639708223 @default.
• W2553100439 hasPublicationYear "2016" @default.
• W2553100439 type Work @default.
• W2553100439 sameAs 2553100439 @default.
• W2553100439 citedByCount "4" @default.
• W2553100439 countsByYear W25531004392018 @default.
• W2553100439 countsByYear W25531004392019 @default.
• W2553100439 countsByYear W25531004392020 @default.
• W2553100439 crossrefType "posted-content" @default.
• W2553100439 hasAuthorship W2553100439A5003929021 @default.
• W2553100439 hasAuthorship W2553100439A5028024287 @default.
• W2553100439 hasAuthorship W2553100439A5032352025 @default.
• W2553100439 hasAuthorship W2553100439A5033588812 @default.
• W2553100439 hasAuthorship W2553100439A5058406424 @default.
• W2553100439 hasAuthorship W2553100439A5090234226 @default.
• W2553100439 hasConcept C104317684 @default.
• W2553100439 hasConcept C108583219 @default.
• W2553100439 hasConcept C113775141 @default.
• W2553100439 hasConcept C11413529 @default.
• W2553100439 hasConcept C121332964 @default.
• W2553100439 hasConcept C153180895 @default.
• W2553100439 hasConcept C154945302 @default.
• W2553100439 hasConcept C155032097 @default.
• W2553100439 hasConcept C163258240 @default.
• W2553100439 hasConcept C17744445 @default.
• W2553100439 hasConcept C185592680 @default.
• W2553100439 hasConcept C193415008 @default.
• W2553100439 hasConcept C199539241 @default.
• W2553100439 hasConcept C204241405 @default.
• W2553100439 hasConcept C2776359362 @default.
• W2553100439 hasConcept C2777735758 @default.
• W2553100439 hasConcept C2944601119 @default.
• W2553100439 hasConcept C31258907 @default.

• next