BROWSE
Detail
Conference Paper
채널 중요도 기반 네트워크 프루닝 방법
Cited - time in 
Share 
Share
- Issue Date
- 2021-07
- Citation
- 대한전자공학회 학술 대회 (하계) 2021, pp.1739-1742
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- In this paper, we propose a channel attention-based network pruning method. Most commonly used in networks, batch normalization learns scale values, which corresponds to channel attention, enabling network pruning. We apply batch normalization with sparsity regularization and combine channel attention module to refine and improve pruning performance. Experimental results show the ability of proposed method, reducing computation and memory footprint while maintaining performance for various networks.
- KSP Keywords
- Batch normalization, Network pruning, Pruning method, Sparsity regularization, memory footprint