Compressing and Accelerating Sparse CNNs Using Sign-Reserved Toeplitz Filters and Input Activation Density-aware Dataflow

doi:10.1145/3773995

CSpace

	Compressing and Accelerating Sparse CNNs Using Sign-Reserved Toeplitz Filters and Input Activation Density-aware Dataflow
	Wang, Zhen 1,2; Liu, Tianyu 1,2; Fan, Zhihua 1,2; Li, Wenming 1,2; Qiu, Yuhang 1,2; Zhang, Zhiyuan 1,2; An, Xuejun 1; Fan, Dongrui 1; Ye, Xiaochun 1
	2025-12-01
发表期刊	ACM TRANSACTIONS ON ARCHITECTURE AND CODE OPTIMIZATION
ISSN	1544-3566
卷号	22 期号:4 页码:23
摘要	Exploiting the sparsity in convolutional neural networks is crucial to accelerate computing and reduce energy consumption. Unstructured sparsity, benefiting from its flexibility to accommodate arbitrary sparse patterns, generally achieves higher accuracy, but it often introduces irregularity in convolutional operations, which complicates the control logic and undermines the benefits of sparsification. Structured sparsity alleviates these problems but sacrifices its application flexibility, which leads to lower accuracy. In this article, we propose TSCNN, an algorithm-hardware co-design solution that aims to compress and accelerate sparse CNNs while balancing both adaptability to sparsity and computational efficiency. In terms of algorithm, TSCNN adopts pruned filters compressed with sign-reserved Toeplitz matrix format (Tfilters), which systematically enhances the regularity of data reuse and flexibly reduces network parameters by 44%-86% while maintaining accuracy. In terms of hardware, TSCNN accelerator adapts to the structure of Tfilters and utilizes density-aware dataflows to support input activations with large sparsity variation, further optimizing the computational efficiency. Experiments show that TSCNN outperforms a dense CNN accelerator, sparse CNN accelerators SCNN and CSCNN, achieving 5.31x, 2.46x and 1.53x speedup and reducing energy consumption by 80.68%, 69.13% and 53.52%, respectively.
关键词	Convolutional neural networks accelerators sparsity algorithm-hardware co-design
DOI	10.1145/3773995
收录类别	SCI
语种	英语
WOS研究方向	Computer Science
WOS类目	Computer Science, Hardware & Architecture ; Computer Science, Theory & Methods
WOS记录号	WOS:001667494000011
出版者	ASSOC COMPUTING MACHINERY
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/42890
专题	中国科学院计算技术研究所
通讯作者	Wang, Zhen
作者单位	1.Chinese Acad Sci, State Key Lab Proc, Inst Comp Technol, Beijing, Peoples R China 2.Univ Chinese Acad Sci, Sch Comp Sci & Technol, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Wang, Zhen,Liu, Tianyu,Fan, Zhihua,et al. Compressing and Accelerating Sparse CNNs Using Sign-Reserved Toeplitz Filters and Input Activation Density-aware Dataflow[J]. ACM TRANSACTIONS ON ARCHITECTURE AND CODE OPTIMIZATION,2025,22(4):23.
APA	Wang, Zhen.,Liu, Tianyu.,Fan, Zhihua.,Li, Wenming.,Qiu, Yuhang.,...&Ye, Xiaochun.(2025).Compressing and Accelerating Sparse CNNs Using Sign-Reserved Toeplitz Filters and Input Activation Density-aware Dataflow.ACM TRANSACTIONS ON ARCHITECTURE AND CODE OPTIMIZATION,22(4),23.
MLA	Wang, Zhen,et al."Compressing and Accelerating Sparse CNNs Using Sign-Reserved Toeplitz Filters and Input Activation Density-aware Dataflow".ACM TRANSACTIONS ON ARCHITECTURE AND CODE OPTIMIZATION 22.4(2025):23.