We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Vision and Pattern Recognition

Title: Anchor-Based Adversarially Robust Zero-Shot Learning Driven by Language

Abstract: Deep neural networks are vulnerable to adversarial attacks. We consider adversarial defense in the case of zero-shot image classification setting, which has rarely been explored because both adversarial defense and zero-shot learning are challenging. We propose LAAT, a novel Language-driven, Anchor-based Adversarial Training strategy, to improve the adversarial robustness in a zero-shot setting. LAAT uses a text encoder to obtain fixed anchors (normalized feature embeddings) of each category, then uses these anchors to perform adversarial training. The text encoder has the property that semantically similar categories can be mapped to neighboring anchors in the feature space. By leveraging this property, LAAT can make the image model adversarially robust on novel categories without any extra examples. Experimental results show that our method achieves impressive zero-shot adversarial performance, even surpassing the previous state-of-the-art adversarially robust one-shot methods in most attacking settings. When models are trained with LAAT on large datasets like ImageNet-1K, they can have substantial zero-shot adversarial robustness across several downstream datasets.
Comments: 11 pages
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as: arXiv:2301.13096 [cs.CV]
  (or arXiv:2301.13096v1 [cs.CV] for this version)

Submission history

From: Xiao Li [view email]
[v1] Mon, 30 Jan 2023 17:34:43 GMT (275kb,D)

Link back to: arXiv, form interface, contact.