Certifiably Robust Neural ODE With Learning-Based Barrier Function

被引：2

作者：

Yang, Runing ^{[1
]}

Jia, Ruoxi ^{[1
]}

Zhang, Xiangyu ^{[2
]}

Jin, Ming ^{[1
]}

机构：

[1] Virginia Tech, Dept Elect & Comp Engn, Blacksburg, VA 24061 USA

[2] Natl Renewable Energy Lab, Computat Sci Ctr, Golden, CO 80401 USA

来源：

IEEE CONTROL SYSTEMS LETTERS | 2023年 / 7卷

关键词：

Robustness; Training; Perturbation methods; Optimization; Government; Standards; Deep learning; Neural networks; machine learning; datadriven control;

D O I：

10.1109/LCSYS.2023.3265397

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

Neural Ordinary Differential Equations (ODEs) have gained traction in many applications. While recent studies have focused on empirically increasing the robustness of neural ODEs against natural or adversarial attacks, certified robustness is still lacking. In this letter, we propose a framework for training a neural ODE using barrier functions and demonstrate improved robustness for classification problems. We further provide the first generalization guarantee of robustness against adversarial attacks using a wait-and-judge scenario approach.

引用

页码：1634 / 1639

页数：6

共 23 条

[1] Ames AD, 2019, 2019 18TH EUROPEAN CONTROL CONFERENCE (ECC), P3420, DOI [10.23919/ECC.2019.8796030, 10.23919/ecc.2019.8796030]
[2] Andriushchenko Maksym, 2020, Computer Vision - ECCV 2020. 16th European Conference. Proceedings. Lecture Notes in Computer Science (LNCS 12368), P484, DOI 10.1007/978-3-030-58592-1_29
[3] Wait-and-judge scenario optimization
Campi, M. C.
Garatti, S.
[J]. MATHEMATICAL PROGRAMMING, 2018, 167 (01) : 155 - 189
[4] Campi MC, 2024, Arxiv, DOI arXiv:2301.12767
[5] A General Scenario Theory for Nonconvex Optimization and Decision Making
Campi, Marco Claudio
Garatti, Simone
Ramponi, Federico Alessandro
[J]. IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 2018, 63 (12) : 4067 - 4078
[6] Chen R. T., 2018, Neural ordinary differential equations
[7] Croce F, 2020, PR MACH LEARN RES, V119
[8] Dawson C, 2022, Arxiv, DOI arXiv:2202.11762
[9] Ghosh A., 2020, Adv. Neural Inf. Process. Syst., V33, P14831
[10] Goodfellow I., 2015, P INT C LEARN REPR

← 1 2 3 →