Keywords: Robustness, Vision Transformer, Invariance
TL;DR: This paper introduces a finetuning approach that improves vision transformer robustness by augmenting training data with synthesized "nullspace noise" that is not expected to affect model outputs.
Abstract: Enhancing the robustness of deep learning models, particularly in the realm of vision transformers (ViTs), is crucial for their real-world deployment. In this work, we provide a finetuning approach to enhance the robustness of vision transformers inspired by the concept of nullspace from linear algebra. Our investigation centers on whether a vision transformer can exhibit resilience to input variations akin to the nullspace property in linear mappings, implying that perturbations sampled from this nullspace do not influence the model's output when added to the input. We start from the observation that many existing ViTs satisfy this property because a non-trivial nullspace exists in their patch embedding layers. Secondly, as nullspace is a concept associated with linear algebra, we demonstrate that it is possible to synthesize approximate nullspace elements for the non-linear blocks of ViTs employing an optimisation strategy. Finally, we propose a fine-tuning strategy for ViTs wherein we augment the training data with synthesized approximate nullspace noise. After finetuning, we find that the model demonstrates robustness to adversarial and natural image perbutations alike.
Submission Number: 77
Loading