Addressing Neural Network Robustness with Mixup and Targeted Labeling Adversarial Training

2020/08/19 by Laugros, Alfred, Caplier, Alice, Ospici, Matthieu
#Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML)

paper · doi:10.48550/arxiv.2008.08384

Abstract

Despite their performance, Artificial Neural Networks are not reliable enough for most of industrial applications. They are sensitive to noises, rotations, blurs and adversarial examples. There is a need to build defenses that protect against a wide range of perturbations, covering the most traditional common corruptions and adversarial examples. We propose a new data augmentation strategy called M-TLAT and designed to address robustness in a broad sense. Our approach combines the Mixup augmentation and a new adversarial training algorithm called Targeted Labeling Adversarial Training (TLAT). The idea of TLAT is to interpolate the target labels of adversarial examples with the ground-truth labels. We show that M-TLAT can increase the robustness of image classifiers towards nineteen common corruptions and five adversarial attacks, without reducing the accuracy on clean samples.

Related

• Masking and Mixing Adversarial Training
• Adversarial Training: embedding adversarial perturbations into the parameter space of a neural network to build a robust system
• A Multiclass Boosting Framework for Achieving Fast and Provable Adversarial Robustness
• Soften to Defend: Towards Adversarial Robustness via Self-Guided Label Refinement
• Enhancing Robust Representation in Adversarial Training: Alignment and Exclusion Criteria