Feature Distillation With Guided Adversarial Contrastive Learning
This addresses the computational expense of adversarial training for deep learning models vulnerable to adversarial attacks, offering an incremental improvement in robustness transfer.
The paper tackles the problem of transferring adversarial robustness from teacher to student models more efficiently than adversarial training, proposing Guided Adversarial Contrastive Distillation (GACD) to use features for this purpose, and achieves comparable or better results on datasets like CIFAR-10, CIFAR-100, and STL-10.
Deep learning models are shown to be vulnerable to adversarial examples. Though adversarial training can enhance model robustness, typical approaches are computationally expensive. Recent works proposed to transfer the robustness to adversarial attacks across different tasks or models with soft labels.Compared to soft labels, feature contains rich semantic information and holds the potential to be applied to different downstream tasks. In this paper, we propose a novel approach called Guided Adversarial Contrastive Distillation (GACD), to effectively transfer adversarial robustness from teacher to student with features. We first formulate this objective as contrastive learning and connect it with mutual information. With a well-trained teacher model as an anchor, students are expected to extract features similar to the teacher. Then considering the potential errors made by teachers, we propose sample reweighted estimation to eliminate the negative effects from teachers. With GACD, the student not only learns to extract robust features, but also captures structural knowledge from the teacher. By extensive experiments evaluating over popular datasets such as CIFAR-10, CIFAR-100 and STL-10, we demonstrate that our approach can effectively transfer robustness across different models and even different tasks, and achieve comparable or better results than existing methods. Besides, we provide a detailed analysis of various methods, showing that students produced by our approach capture more structural knowledge from teachers and learn more robust features under adversarial attacks.