Repository logo
 

Improving the robustness of analog deep neural networks through a Bayes-optimized noise injection approach

Published version
Peer-reviewed

Repository DOI


Change log

Authors

Cao, Linfeng 
Yang, Liujia 
Zhang, Ziqing 
Fang, Zhicheng 

Abstract

jats:titleAbstract</jats:title>jats:pAnalog deep neural networks (DNNs) provide a promising solution, especially for deployment on resource-limited platforms, for example in mobile settings. However, the practicability of analog DNNs has been limited by their instability due to multi-factor reasons from manufacturing, thermal noise, etc. Here, we present a theoretically guaranteed noise injection approach to improve the robustness of analog DNNs without any hardware modifications or sacrifice of accuracy, which proves that within a certain range of parameter perturbations, the prediction results would not change. Experimental results demonstrate that our algorithmic framework can outperform state-of-the-art methods on tasks including image classification, object detection, and large-scale point cloud object detection in autonomous driving by a factor of 10 to 100. Together, our results may serve as a way to ensure the robustness of analog deep neural network systems, especially for safety-critical applications.</jats:p>

Description

Acknowledgements: N.Y. acknowledges the funding from the National Science Foundation of China (Grant No. 62106139). Q.G. is grateful for the support from the Shanghai Artificial Intelligence Laboratory and the National Key R&D Program of China (Grant NO.2022ZD0160100). G.-Z.Y. acknowledges the funding from the Science and Technology Commission of Shanghai Municipality (Grant No. 20DZ2220400).

Keywords

46 Information and Computing Sciences, 4611 Machine Learning, Machine Learning and Artificial Intelligence, Bioengineering

Journal Title

Communications Engineering

Conference Name

Journal ISSN

2731-3395
2731-3395

Volume Title

Publisher

Springer Science and Business Media LLC
Sponsorship
National Natural Science Foundation of China (National Science Foundation of China) (62106139)