Journal of Computer Research and Development
|
2025, 62(2): 532-541
• High Performance Computing •
KCNN: A Neural Network Lightweight Method and Hardware Implementation Architecture
Full
Guilin Chen
, Guanwu Wang*
, Kang Wang, Minhui Hu, Junquan Deng
Affiliations
- The 63rd Research Institute, National University of Defense Technology, Nanjing 210007
born in 1994. PhD candidate, engineer. His main interests include architecture, and Chiplet and neural network accelerator. (cglnudt@163.com)
born in 1986. PhD, assistant professor. His main research interests include architecture and system-on-chip design
born in 1990. Master, assistant professor. His main research interests include architecture and system-on-chip design
born in 1994. PhD. Her main research interests include artificial intelligence and computer architecture
born in 1988. PhD, associate research fellow. His main research interest includes radio resource management, mobile positioning, and machine learning in wireless networks
About Author:
born in 1994. PhD candidate, engineer. His main interests include architecture, and Chiplet and neural network accelerator. (cglnudt@163.com)
born in 1986. PhD, assistant professor. His main research interests include architecture and system-on-chip design
born in 1990. Master, assistant professor. His main research interests include architecture and system-on-chip design
born in 1994. PhD. Her main research interests include artificial intelligence and computer architecture
born in 1988. PhD, associate research fellow. His main research interest includes radio resource management, mobile positioning, and machine learning in wireless networks
Published: 2025-02-01
doi: 10.7544/issn1000-1239.202330409
Outline
Convolutional neural network (CNN) has become one of the most important machine learning technologies in the field of image recognition. In recent years, with the increasing demand for CNN deployment at the mobile edge, the lightweight of CNN has become a research hotspot. The mainstream CNN lightweight methods include pruning and quantization, both of which can effectively reduce the calculation and storage overhead for CNN inference. However, none of these methods fully exploits the bilateral sparsity (weight sparsity and activation sparsity) and potential data-reuse in CNN. Therefore, in order to solve these problems, we propose a new neural network lightweight method, and the k-means algorithm is used to cluster the non-zero values of convolution kernel and feature map, and the CNN inference only uses limited cluster values as multipliers to complete all convolutional calculations. Compared with the computational complexity of original convolutional layer $ O({n}^{3}) $, the computational complexity of the convolutional layer after lightweight is $ O({n}^{2}) $, effectively reducing the amount of computation. Similarly, the non-zero weights of the fully connected layer are also clustered, and only the cluster values and corresponding index vectors are stored on chip, which significantly reduces the storage overhead. Finally, a customized architecture KCNN is designed for this lightweight method. The architecture modularizes the different processes of the CNN, and compared with the previous accelerator, a non-zero clustering module is added. In addition, some caches are added to make use of the data-reuse in the CNN after clustering. The experiment results show that without losing the derivation accuracy, the overall calculation of AlexNet network is reduced by 66%, and the storage expense is reduced by 85%.
k-means
/
CNN
/
lightweight
/
acceleration
/
compression
陈桂林, 王观武, 王康, 胡敏慧, 邓俊荃.
KCNN:一种神经网络轻量化方法和硬件实现架构.
计算机研究与发展,
2025
, 62
(2)
: 532
-541
.
DOI: 10.7544/issn1000-1239.202330409
Guilin Chen, Guanwu Wang, Kang Wang, Minhui Hu, Junquan Deng.
KCNN: A Neural Network Lightweight Method and Hardware Implementation Architecture[J].
Journal of Computer Research and Development,
2025
, 62
(2)
: 532
-541
.
DOI: 10.7544/issn1000-1239.202330409
Year 2025 volume 62 Issue 2
99
49
Cite this Article
BibTeX
Article Info
doi: 10.7544/issn1000-1239.202330409
- Receive Date:2023-05-26
- Online Date:2026-11-10
- Published:2025-02-01