Similarity‐based adversarial knowledge distillation using graph convolutional neural network
Abstract This letter presents an adversarial knowledge distillation based on graph convolutional neural network. For knowledge distillation, many methods have been proposed in which the student model individually and independently imitates the output of the teacher model on the input data. Our metho...
Published in: | Electronics Letters |
---|---|
Main Authors: | , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
Wiley
2022
|
Subjects: | |
Online Access: | https://doi.org/10.1049/ell2.12543 https://doaj.org/article/9c39d2c5854c435ca81726a22f5637c9 |
id |
ftdoajarticles:oai:doaj.org/article:9c39d2c5854c435ca81726a22f5637c9 |
---|---|
record_format |
openpolar |
spelling |
ftdoajarticles:oai:doaj.org/article:9c39d2c5854c435ca81726a22f5637c9 2023-05-15T16:01:51+02:00 Similarity‐based adversarial knowledge distillation using graph convolutional neural network Sungjun Lee Sejun Kim Seong Soo Kim Kisung Seo 2022-08-01T00:00:00Z https://doi.org/10.1049/ell2.12543 https://doaj.org/article/9c39d2c5854c435ca81726a22f5637c9 EN eng Wiley https://doi.org/10.1049/ell2.12543 https://doaj.org/toc/0013-5194 https://doaj.org/toc/1350-911X 1350-911X 0013-5194 doi:10.1049/ell2.12543 https://doaj.org/article/9c39d2c5854c435ca81726a22f5637c9 Electronics Letters, Vol 58, Iss 16, Pp 606-608 (2022) Electrical engineering. Electronics. Nuclear engineering TK1-9971 article 2022 ftdoajarticles https://doi.org/10.1049/ell2.12543 2022-12-31T00:57:34Z Abstract This letter presents an adversarial knowledge distillation based on graph convolutional neural network. For knowledge distillation, many methods have been proposed in which the student model individually and independently imitates the output of the teacher model on the input data. Our method suggests the application of a similarity matrix to consider the relationship among output vectors, compared to the other existing approaches. The similarity matrix of the output vectors is calculated and converted into a graph structure, and a generative adversarial network using graph convolutional neural network is applied. We suggest similarity‐based knowledge distillation in which a student model simultaneously imitates both of output vector and similarity matrix of the teacher model. We evaluate our method on ResNet, MobileNet and Wide ResNet using CIFAR‐10 and CIFAR‐100 datasets, and our results outperform results of the baseline model and other existing knowledge distillations like KLD and DML. Article in Journal/Newspaper DML Directory of Open Access Journals: DOAJ Articles Electronics Letters 58 16 606 608 |
institution |
Open Polar |
collection |
Directory of Open Access Journals: DOAJ Articles |
op_collection_id |
ftdoajarticles |
language |
English |
topic |
Electrical engineering. Electronics. Nuclear engineering TK1-9971 |
spellingShingle |
Electrical engineering. Electronics. Nuclear engineering TK1-9971 Sungjun Lee Sejun Kim Seong Soo Kim Kisung Seo Similarity‐based adversarial knowledge distillation using graph convolutional neural network |
topic_facet |
Electrical engineering. Electronics. Nuclear engineering TK1-9971 |
description |
Abstract This letter presents an adversarial knowledge distillation based on graph convolutional neural network. For knowledge distillation, many methods have been proposed in which the student model individually and independently imitates the output of the teacher model on the input data. Our method suggests the application of a similarity matrix to consider the relationship among output vectors, compared to the other existing approaches. The similarity matrix of the output vectors is calculated and converted into a graph structure, and a generative adversarial network using graph convolutional neural network is applied. We suggest similarity‐based knowledge distillation in which a student model simultaneously imitates both of output vector and similarity matrix of the teacher model. We evaluate our method on ResNet, MobileNet and Wide ResNet using CIFAR‐10 and CIFAR‐100 datasets, and our results outperform results of the baseline model and other existing knowledge distillations like KLD and DML. |
format |
Article in Journal/Newspaper |
author |
Sungjun Lee Sejun Kim Seong Soo Kim Kisung Seo |
author_facet |
Sungjun Lee Sejun Kim Seong Soo Kim Kisung Seo |
author_sort |
Sungjun Lee |
title |
Similarity‐based adversarial knowledge distillation using graph convolutional neural network |
title_short |
Similarity‐based adversarial knowledge distillation using graph convolutional neural network |
title_full |
Similarity‐based adversarial knowledge distillation using graph convolutional neural network |
title_fullStr |
Similarity‐based adversarial knowledge distillation using graph convolutional neural network |
title_full_unstemmed |
Similarity‐based adversarial knowledge distillation using graph convolutional neural network |
title_sort |
similarity‐based adversarial knowledge distillation using graph convolutional neural network |
publisher |
Wiley |
publishDate |
2022 |
url |
https://doi.org/10.1049/ell2.12543 https://doaj.org/article/9c39d2c5854c435ca81726a22f5637c9 |
genre |
DML |
genre_facet |
DML |
op_source |
Electronics Letters, Vol 58, Iss 16, Pp 606-608 (2022) |
op_relation |
https://doi.org/10.1049/ell2.12543 https://doaj.org/toc/0013-5194 https://doaj.org/toc/1350-911X 1350-911X 0013-5194 doi:10.1049/ell2.12543 https://doaj.org/article/9c39d2c5854c435ca81726a22f5637c9 |
op_doi |
https://doi.org/10.1049/ell2.12543 |
container_title |
Electronics Letters |
container_volume |
58 |
container_issue |
16 |
container_start_page |
606 |
op_container_end_page |
608 |
_version_ |
1766397556774928384 |