Loading...
Loading...

Go to the content (press return)

The dark side of DNN pruning

Author
Yazdani, R.; Riera, M.; Arnau, J.; Gonzalez, A.
Type of activity
Presentation of work at congresses
Name of edition
The 45th International Symposium on Computer Architecture
Date of publication
2018
Presentation's date
2018-06-02
Book of congress proceedings
2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA 2018): Los Angeles, California, USA: 1-6 June 2018
First page
790
Last page
801
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
DOI
https://doi.org/10.1109/ISCA.2018.00071 Open in new window
Project funding
Intelligent, Ubiquitous and Energy-Efficient Computing Systems
Repository
http://hdl.handle.net/2117/125141 Open in new window
URL
https://ieeexplore.ieee.org/document/8416873 Open in new window
Abstract
DNN pruning has been recently proposed as an effective technique to improve the energy-efficiency of DNN-based solutions. It is claimed that by removing unimportant or redundant connections, the pruned DNN delivers higher performance and energy-efficiency with negligible impact on accuracy. However, DNN pruning has an important side effect: it May reduce the confidence of DNN predictions. We show that, although top-1 accuracy May be maintained with DNN pruning, the likelihood of the class in the...
Citation
Yazdani, R., Riera, M., Arnau, J., Gonzalez Colas, A. The dark side of DNN pruning. A: International Symposium on Computer Architecture. "2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA 2018): Los Angeles, California, USA: 1-6 June 2018". Institute of Electrical and Electronics Engineers (IEEE), 2018, p. 790-801.
Keywords
Automatic speech recognition, Automatic speech recognition (ASR), Computer architecture, Computer hardware, DNN pruning, Deep learning, Deep neural networks, Energy conservation, Energy efficiency, Hardware, Hardware accelerator, Hardware accelerators, N-best hypothesis, Redundant connections, Solution approach, Speech recognition, State of the art, Viterbi algorithm, Viterbi search
Group of research
ARCO - Microarchitecture and Compilers

Participants