Online Program

Return to main conference page
Friday, May 31
Machine Learning
Recent Advancements in Deep Learning
Fri, May 31, 5:20 PM - 6:25 PM
Regency Ballroom AB

Optimal Transport Classifier: Defending Against Adversarial Attacks by Regularized Deep Embedding (306208)

Cho-Jui Hsieh, University of California, Los Angeles 
Erik Kruus, NEC Laboratories America, Inc. 
Thomas Lee, UC Davis 
*Yao Li, University of California, Davis 
Martin Renqiang Min, NEC Laboratories America, Inc. 
Wenchao Yu, University of California, Los Angeles 

Keywords: Machine learning, Deep learning, Computer Vision, Adversarial Examples, Model Robustness

Recent studies have demonstrated the vulnerability of deep convolutional neural networks against adversarial examples. Inspired by the observation that the intrinsic dimension of image data is much smaller than its pixel space dimension and the vulnerability of neural networks grows with the input dimension, we propose to embed high-dimensional input images into a low-dimensional space to perform classification. However, arbitrarily projecting the input images to a low-dimensional space without regularization will not improve the robustness of deep neural networks. Leveraging optimal transport theory, we propose a new framework, Optimal Transport Classifier (OT-Classifier), and derive an objective that minimizes the discrepancy between the distribution of the true label and the distribution of the OT-Classifier output. Experimental results on several benchmark datasets show that, our proposed framework achieves state-of-the-art performance against strong adversarial attack methods.