Matches in SemOpenAlex for { <https://semopenalex.org/work/W3195881387> ?p ?o ?g. }
- W3195881387 endingPage "115314" @default.
- W3195881387 startingPage "115292" @default.
- W3195881387 abstract "Convolutional neural networks (CNNs) have developed to become powerful models for various computer vision tasks ranging from object detection to semantic segmentation. However, most of the state-of-the-art CNNs cannot be deployed directly on edge devices such as smartphones and drones, which need low latency under limited power and memory bandwidth. One popular, straightforward approach to compressing CNNs is network slimming, which imposes $ell _{1}$ regularization on the channel-associated scaling factors via the batch normalization layers during training. Network slimming thereby identifies insignificant channels that can be pruned for inference. In this paper, we propose replacing the $ell _{1}$ penalty with an alternative nonconvex, sparsity-inducing penalty in order to yield a more compressed and/or accurate CNN architecture. We investigate $ell _{p} (0 , transformed $ell _{1}$ ( $text{T}ell _{1}$ ), minimax concave penalty (MCP), and smoothly clipped absolute deviation (SCAD) due to their recent successes and popularity in solving sparse optimization problems, such as compressed sensing and variable selection. We demonstrate the effectiveness of network slimming with nonconvex penalties on three neural network architectures – VGG-19, DenseNet-40, and ResNet-164 – on standard image classification datasets. Based on the numerical experiments, $text{T}ell _{1}$ preserves model accuracy against channel pruning, $ell _{1/2, 3/4}$ yield better compressed models with similar accuracies after retraining as $ell _{1}$ , and MCP and SCAD provide more accurate models after retraining with similar compression as $ell _{1}$ . Network slimming with $text{T}ell _{1}$ regularization also outperforms the latest Bayesian modification of network slimming in compressing a CNN architecture in terms of memory storage while preserving its model accuracy after channel pruning." @default.
- W3195881387 created "2021-08-30" @default.
- W3195881387 creator A5007322337 @default.
- W3195881387 creator A5016173634 @default.
- W3195881387 creator A5030190007 @default.
- W3195881387 creator A5059708262 @default.
- W3195881387 creator A5079155274 @default.
- W3195881387 date "2021-01-01" @default.
- W3195881387 modified "2023-10-18" @default.
- W3195881387 title "Improving Network Slimming With Nonconvex Regularization" @default.
- W3195881387 cites W1165639838 @default.
- W3195881387 cites W1677182931 @default.
- W3195881387 cites W1901129140 @default.
- W3195881387 cites W1903029394 @default.
- W3195881387 cites W1908384797 @default.
- W3195881387 cites W1943431218 @default.
- W3195881387 cites W1965125844 @default.
- W3195881387 cites W1972163814 @default.
- W3195881387 cites W1991857131 @default.
- W3195881387 cites W1996287810 @default.
- W3195881387 cites W1996901117 @default.
- W3195881387 cites W2004544971 @default.
- W3195881387 cites W2014360396 @default.
- W3195881387 cites W2025666718 @default.
- W3195881387 cites W2032944446 @default.
- W3195881387 cites W2056201402 @default.
- W3195881387 cites W2074682976 @default.
- W3195881387 cites W2089106993 @default.
- W3195881387 cites W2101675075 @default.
- W3195881387 cites W2102605133 @default.
- W3195881387 cites W2138019504 @default.
- W3195881387 cites W2138049103 @default.
- W3195881387 cites W2145096794 @default.
- W3195881387 cites W2155024610 @default.
- W3195881387 cites W2158928439 @default.
- W3195881387 cites W2164452299 @default.
- W3195881387 cites W2168745297 @default.
- W3195881387 cites W2183341477 @default.
- W3195881387 cites W2194775991 @default.
- W3195881387 cites W2221775387 @default.
- W3195881387 cites W2331143823 @default.
- W3195881387 cites W2412782625 @default.
- W3195881387 cites W2460144244 @default.
- W3195881387 cites W2557728737 @default.
- W3195881387 cites W2605135468 @default.
- W3195881387 cites W2743133559 @default.
- W3195881387 cites W2763531044 @default.
- W3195881387 cites W2886697509 @default.
- W3195881387 cites W2888333721 @default.
- W3195881387 cites W2896556344 @default.
- W3195881387 cites W2962689221 @default.
- W3195881387 cites W2962851801 @default.
- W3195881387 cites W2962958489 @default.
- W3195881387 cites W2963125010 @default.
- W3195881387 cites W2963163009 @default.
- W3195881387 cites W2963446712 @default.
- W3195881387 cites W2964010532 @default.
- W3195881387 cites W2964088520 @default.
- W3195881387 cites W2964266063 @default.
- W3195881387 cites W2964461714 @default.
- W3195881387 cites W2970738028 @default.
- W3195881387 cites W3008070224 @default.
- W3195881387 cites W3033890399 @default.
- W3195881387 cites W3034818206 @default.
- W3195881387 cites W3100817920 @default.
- W3195881387 cites W3106108064 @default.
- W3195881387 cites W3129991176 @default.
- W3195881387 cites W3135817118 @default.
- W3195881387 cites W317954863 @default.
- W3195881387 cites W3182242813 @default.
- W3195881387 cites W4250482878 @default.
- W3195881387 doi "https://doi.org/10.1109/access.2021.3105366" @default.
- W3195881387 hasPublicationYear "2021" @default.
- W3195881387 type Work @default.
- W3195881387 sameAs 3195881387 @default.
- W3195881387 citedByCount "3" @default.
- W3195881387 countsByYear W31958813872021 @default.
- W3195881387 countsByYear W31958813872022 @default.
- W3195881387 countsByYear W31958813872023 @default.
- W3195881387 crossrefType "journal-article" @default.
- W3195881387 hasAuthorship W3195881387A5007322337 @default.
- W3195881387 hasAuthorship W3195881387A5016173634 @default.
- W3195881387 hasAuthorship W3195881387A5030190007 @default.
- W3195881387 hasAuthorship W3195881387A5059708262 @default.
- W3195881387 hasAuthorship W3195881387A5079155274 @default.
- W3195881387 hasBestOaLocation W31958813871 @default.
- W3195881387 hasConcept C11413529 @default.
- W3195881387 hasConcept C154945302 @default.
- W3195881387 hasConcept C2776135515 @default.
- W3195881387 hasConcept C41008148 @default.
- W3195881387 hasConceptScore W3195881387C11413529 @default.
- W3195881387 hasConceptScore W3195881387C154945302 @default.
- W3195881387 hasConceptScore W3195881387C2776135515 @default.
- W3195881387 hasConceptScore W3195881387C41008148 @default.
- W3195881387 hasFunder F4320306076 @default.
- W3195881387 hasLocation W31958813871 @default.
- W3195881387 hasLocation W31958813872 @default.
- W3195881387 hasOpenAccess W3195881387 @default.