Robust Representation Learning via Asymmetric Negative Contrasting and Reverse Attention

Nuoyan Zhou, Decheng Liu, Dawei Zhou, Nannan Wang, Chunlei Peng, Xinbo Gao

Keywords: Robust Representation learning, Asymmetric Negative Contrast, Reverse Attention

TL;DR: Learning robust feature makes AT great again

Abstract:

Deep neural networks are vulnerable to adversarial noise. Adversarial training (AT) has been demonstrated to be the most effective defense strategy to protect neural networks from being fooled. However, we find AT omits to learning robust features, resulting in poor performance of adversarial robustness. To address this issue, we highlight two characteristics of robust representation:
(1) exclusion: feature of natural examples keeps away from that of other classes;
(2) alignment: feature of natural and corresponding adversarial examples is close to each other.
These motivate us to propose a generic framework of AT to gain robust representation, by the asymmetric negative contrast and reverse attention. Specifically, we design an asymmetric negative contrast based on predicted probabilities and generate adversarial negative examples by the targeted attack, to push away examples of different classes in the feature space. Moreover, we propose to weight feature by parameters of the linear classifier as the reverse attention, to obtain class-aware feature and pull close the feature of the same class. Empirical evaluations on three benchmark datasets show our method greatly advances the robustness of AT and achieves the state-of-the-art performance.

Submission Number: 5667