Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness

Li Xiao; Yijie Peng; L. Jeff Hong; Zewu Ke; Shuhuai Yang

doi:10.1109/CASE48305.2020.9216839

Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness

Li Xiao, Yijie Peng, L. Jeff Hong, Zewu Ke, Shuhuai Yang

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

2 Scopus citations

Abstract

In this work, we proposed a generalized likelihood ratio method capable of training the artificial neural networks with more flexibility: (a)training with discrete activation and loss functions, while the traditional back propagation method cannot train the artificial neural networks with such activations and loss; (b)involving neuronal noises during training and prediction, which will improve the freedom of the model and make it more adaptable to the real environment, especially when environmental noises exist. Numerical results show that the robustness of various artificial neural networks trained by the new method is significantly improved when the input data is affected by both the natural noises and adversarial attacks.

Original language	English (US)
Title of host publication	2020 IEEE 16th International Conference on Automation Science and Engineering, CASE 2020
Publisher	IEEE Computer Society
Pages	1343-1348
Number of pages	6
ISBN (Electronic)	9781728169040
DOIs	https://doi.org/10.1109/CASE48305.2020.9216839
State	Published - Aug 2020
Externally published	Yes
Event	16th IEEE International Conference on Automation Science and Engineering, CASE 2020 - Hong Kong, Hong Kong Duration: Aug 20 2020 → Aug 21 2020

Publication series

Name	IEEE International Conference on Automation Science and Engineering
Volume	2020-August
ISSN (Print)	2161-8070
ISSN (Electronic)	2161-8089

Conference

Conference	16th IEEE International Conference on Automation Science and Engineering, CASE 2020
Country/Territory	Hong Kong
City	Hong Kong
Period	8/20/20 → 8/21/20

Bibliographical note

Publisher Copyright:
© 2020 IEEE.

Publisher link

10.1109/CASE48305.2020.9216839

Find It

Find It at U of M Libraries

Cite this

Xiao, L., Peng, Y., Jeff Hong, L., Ke, Z., & Yang, S. (2020). Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness. In 2020 IEEE 16th International Conference on Automation Science and Engineering, CASE 2020 (pp. 1343-1348). Article 9216839 (IEEE International Conference on Automation Science and Engineering; Vol. 2020-August). IEEE Computer Society. https://doi.org/10.1109/CASE48305.2020.9216839

Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness. / Xiao, Li; Peng, Yijie; Jeff Hong, L. et al.
2020 IEEE 16th International Conference on Automation Science and Engineering, CASE 2020. IEEE Computer Society, 2020. p. 1343-1348 9216839 (IEEE International Conference on Automation Science and Engineering; Vol. 2020-August).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Xiao, L, Peng, Y, Jeff Hong, L, Ke, Z & Yang, S 2020, Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness. in 2020 IEEE 16th International Conference on Automation Science and Engineering, CASE 2020., 9216839, IEEE International Conference on Automation Science and Engineering, vol. 2020-August, IEEE Computer Society, pp. 1343-1348, 16th IEEE International Conference on Automation Science and Engineering, CASE 2020, Hong Kong, Hong Kong, 8/20/20. https://doi.org/10.1109/CASE48305.2020.9216839

Xiao L, Peng Y, Jeff Hong L, Ke Z, Yang S. Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness. In 2020 IEEE 16th International Conference on Automation Science and Engineering, CASE 2020. IEEE Computer Society. 2020. p. 1343-1348. 9216839. (IEEE International Conference on Automation Science and Engineering). doi: 10.1109/CASE48305.2020.9216839

@inproceedings{b90ce2e4f6e14b13bdab1024db3ded89,

title = "Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness",

abstract = "In this work, we proposed a generalized likelihood ratio method capable of training the artificial neural networks with more flexibility: (a)training with discrete activation and loss functions, while the traditional back propagation method cannot train the artificial neural networks with such activations and loss; (b)involving neuronal noises during training and prediction, which will improve the freedom of the model and make it more adaptable to the real environment, especially when environmental noises exist. Numerical results show that the robustness of various artificial neural networks trained by the new method is significantly improved when the input data is affected by both the natural noises and adversarial attacks.",

author = "Li Xiao and Yijie Peng and {Jeff Hong}, L. and Zewu Ke and Shuhuai Yang",

note = "Publisher Copyright: {\textcopyright} 2020 IEEE.; 16th IEEE International Conference on Automation Science and Engineering, CASE 2020 ; Conference date: 20-08-2020 Through 21-08-2020",

year = "2020",

month = aug,

doi = "10.1109/CASE48305.2020.9216839",

language = "English (US)",

series = "IEEE International Conference on Automation Science and Engineering",

publisher = "IEEE Computer Society",

pages = "1343--1348",

booktitle = "2020 IEEE 16th International Conference on Automation Science and Engineering, CASE 2020",

}

TY - GEN

T1 - Training Artificial Neural Networks by Generalized Likelihood Ratio Method

T2 - 16th IEEE International Conference on Automation Science and Engineering, CASE 2020

AU - Xiao, Li

AU - Peng, Yijie

AU - Jeff Hong, L.

AU - Ke, Zewu

AU - Yang, Shuhuai

PY - 2020/8

Y1 - 2020/8

N2 - In this work, we proposed a generalized likelihood ratio method capable of training the artificial neural networks with more flexibility: (a)training with discrete activation and loss functions, while the traditional back propagation method cannot train the artificial neural networks with such activations and loss; (b)involving neuronal noises during training and prediction, which will improve the freedom of the model and make it more adaptable to the real environment, especially when environmental noises exist. Numerical results show that the robustness of various artificial neural networks trained by the new method is significantly improved when the input data is affected by both the natural noises and adversarial attacks.

AB - In this work, we proposed a generalized likelihood ratio method capable of training the artificial neural networks with more flexibility: (a)training with discrete activation and loss functions, while the traditional back propagation method cannot train the artificial neural networks with such activations and loss; (b)involving neuronal noises during training and prediction, which will improve the freedom of the model and make it more adaptable to the real environment, especially when environmental noises exist. Numerical results show that the robustness of various artificial neural networks trained by the new method is significantly improved when the input data is affected by both the natural noises and adversarial attacks.

UR - http://www.scopus.com/inward/record.url?scp=85094098432&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85094098432&partnerID=8YFLogxK

U2 - 10.1109/CASE48305.2020.9216839

DO - 10.1109/CASE48305.2020.9216839

M3 - Conference contribution

AN - SCOPUS:85094098432

T3 - IEEE International Conference on Automation Science and Engineering

SP - 1343

EP - 1348

BT - 2020 IEEE 16th International Conference on Automation Science and Engineering, CASE 2020

PB - IEEE Computer Society

Y2 - 20 August 2020 through 21 August 2020

ER -

Training Artificial Neural Networks by Generalized Likelihood Ratio Method: An Effective Way to Improve Robustness

Abstract

Publication series

Conference

Bibliographical note

Publisher link

Find It

Other files and links

Fingerprint

Cite this