TY - GEN
T1 - BFConv
T2 - 28th International Conference on Neural Information Processing, ICONIP 2021
AU - Yang, Dengjie
AU - Yu, Xuehui
AU - Sun, Yi
AU - Zhuang, Fuzhen
AU - He, Qing
AU - Ye, Shiwei
N1 - Publisher Copyright:
© 2021, Springer Nature Switzerland AG.
PY - 2021
Y1 - 2021
N2 - Convolutional neural network (CNN) is a basic neural network widely used in vision tasks. Many CNNs alleviate the redundancy in feature maps to reduce model complexity. Inspired by digital signal processing theories, this paper reviews discrete fourier transform (DFT), finding its similarities with standard convolution. In particular, DFT has a fast algorithm called FFT, which sparks our thinking: can we learn from the idea of FFT to realize a more efficient convolution filter? Based on the butterfly operation of FFT, we propose a novel butterfly convolution (BFConv). In addition, we illustrate that group weight sharing convolution is a basic unit of BFConv. Compared with the traditional group convolution structure, BFConv constructs group residual-like connections and increases the range of receptive fields for each sub-feature layer. Without changing the network architecture, we integrate BFConv into ResNet-50, ShuffleNet and VGG-16. Experimental results on CIFAR-10 and ImageNet demonstrate the above BFConv-equipped networks reduce parameters and computation, achieving similar or higher accuracy. Remarkably, when ResNet-50 embedded BFConv reaches nearly half of the compression ratio of the model, it performs favorably against its state-of-the-art competitors.
AB - Convolutional neural network (CNN) is a basic neural network widely used in vision tasks. Many CNNs alleviate the redundancy in feature maps to reduce model complexity. Inspired by digital signal processing theories, this paper reviews discrete fourier transform (DFT), finding its similarities with standard convolution. In particular, DFT has a fast algorithm called FFT, which sparks our thinking: can we learn from the idea of FFT to realize a more efficient convolution filter? Based on the butterfly operation of FFT, we propose a novel butterfly convolution (BFConv). In addition, we illustrate that group weight sharing convolution is a basic unit of BFConv. Compared with the traditional group convolution structure, BFConv constructs group residual-like connections and increases the range of receptive fields for each sub-feature layer. Without changing the network architecture, we integrate BFConv into ResNet-50, ShuffleNet and VGG-16. Experimental results on CIFAR-10 and ImageNet demonstrate the above BFConv-equipped networks reduce parameters and computation, achieving similar or higher accuracy. Remarkably, when ResNet-50 embedded BFConv reaches nearly half of the compression ratio of the model, it performs favorably against its state-of-the-art competitors.
KW - Butterfly convolution
KW - Convolutional neural network
KW - FFT
KW - Group Weight sharing convolution
UR - https://www.scopus.com/pages/publications/85121927514
U2 - 10.1007/978-3-030-92273-3_4
DO - 10.1007/978-3-030-92273-3_4
M3 - 会议稿件
AN - SCOPUS:85121927514
SN - 9783030922726
T3 - Lecture Notes in Computer Science
SP - 40
EP - 50
BT - Neural Information Processing - 28th International Conference, ICONIP 2021, Proceedings
A2 - Mantoro, Teddy
A2 - Lee, Minho
A2 - Ayu, Media Anugerah
A2 - Wong, Kok Wai
A2 - Hidayanto, Achmad Nizar
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 8 December 2021 through 12 December 2021
ER -