arrow
返回

High-precision linearized interpretation for fully connected neural network

delete2021-09-01
delete17
PRE
AI
X
Xia Lei
Y
Yongkai Fan *
K
Kuan‐Ching Li
A
Arcangelo Castiglione
Q
Qian Hu
DOI:10.1016/j.asoc.2021.107572delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
Despite the widespread application of deep neural networks in finance, medical treatment, and autonomous driving, these networks face multiple security threats, such as maliciously constructed adversarial samples that can easily mislead deep neural network model classification, causing errors. Therefore, creating an interpretable model or designing an interpretation method is necessary to improve its security. This paper presents an interpretation scheme, named Convergent Interpretation for Deep Neural Networks (CIDNN), to obtain a provably convergent and consistent interpretation for deep neural networks. The main idea of CIDNN is to first convert the deep neural networks into a set of mathematically convergent Piecewise Linear Neural Networks (PLNN), then convert the PLNN into a set of equivalent linear classifiers. In this way, each linear classifier can be interpreted by its decision features. By analyzing the convergence of the local approximation interpretation scheme, we prove that this interpretable model can be sufficiently close to the deep neural network with certain conditions. Experiments show the convergence of CIDNN's interpretation, and the interpretation conforms with similar samples in the synthetic dataset. Besides, we demonstrate the semantical meaning of CIDNN in the Fashion-MNIST dataset. (C) 2021 Elsevier B.V. All rights reserved.
Keyword:
Deep neural network
Convergent interpretation
Piecewise linear function
Sigmoid function
AI总结

AI总结

对已上传原文的论文进行重点信息的提取,主要内容包括:简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。

期刊

Applied Soft Computing 封面图
Applied Soft Computing
IF:
6.6
论文数:
1.4W
被引数:
4.8W

机构

U
University of Salerno
学者数:
1.2W
论文数: 1.1W
被引数: 1.2W
C
Communication University of China
学者数:
1.1K
论文数: 820
被引数: 326
P
providence university - taiwan
学者数:
842
论文数: 1.0K
被引数: 0
C
china university of petroleum
学者数:
4.1W
论文数: 2.7W
被引数: 30
学者 查看更多机构
引用论文

引用论文

err分享
err收藏
Sparse additive models
err2009-10-19
err423
errOAAI
errRavikumar, Pradeep; Lafferty, John; Liu, Han; Wasserman, Larry
err分享
err收藏
One secure data integrity verification scheme for cloud storage
err2019-07-01
err57
PREAI
errFan, Yongkai; Lin, Xiaodong; Tan, Gang; Zhang, Yuqing; Dong, Wei; Lei, Jing
err分享
err收藏
Privacy preserving based logistic regression on big data大数据环境下基于logistic回归的隐私保护
err2020-12-01
err48
PREAI
errFan, Yongkai; Bai, Jianrong; Lei, Xia; Zhang, Yuqing; Zhang, Bin; Li, Kuan-Ching; Tan, Gang
err分享
err收藏
SBBS: A Secure Blockchain-Based Scheme for IoT Data Credibility in Fog EnvironmentSBBS: 雾环境下基于区块链的物联网数据可信性安全方案
err2021-06-01
err25
PREAI
errFan, Yongkai; Zhao, Guanqun; Lei, Xia; Liang, Wei; Li, Kuan-Ching; Choo, Kim-Kwang Raymond; Zhu, Chunsheng
err分享
err收藏
Modelling ice-cliff backwasting on a debris-covered glacier in the Nepalese Himalaya
err2017-07-10
err0
errOAAI
errJakob F. Steiner; Francesca Pellicciotti; Pascal Buri; Evan S. Miles; Walter W. Immerzeel; Tim D. Reid
err分享
err收藏
Predicting cancer outcomes from histology and genomics using convolutional networks使用卷积网络从组织学和基因组学预测癌症结果
err2018-03-12
err683
errOAAI
errMobadersany, Pooya; Yousefi, Safoora; Amgad, Mohamed; Gutman, David A.; Barnholtz-Sloan, Jill S.; Vega, Jose E. Velazquez; Brat, Daniel J.; Cooper, Lee A. D.
err分享
err收藏
没有更多内容