Improving Interpretation Faithfulness for Vision Transformers

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Hu, Lijie, Liu, Yixin, Liu, Ninghao, Huai, Mengdi, Sun, Lichao, Wang, Di
Formato: Preprint
Publicado: 2023
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911864092360704
author Hu, Lijie
Liu, Yixin
Liu, Ninghao
Huai, Mengdi
Sun, Lichao
Wang, Di
author_facet Hu, Lijie
Liu, Yixin
Liu, Ninghao
Huai, Mengdi
Sun, Lichao
Wang, Di
contents Vision Transformers (ViTs) have achieved state-of-the-art performance for various vision tasks. One reason behind the success lies in their ability to provide plausible innate explanations for the behavior of neural architectures. However, ViTs suffer from issues with explanation faithfulness, as their focal points are fragile to adversarial attacks and can be easily changed with even slight perturbations on the input image. In this paper, we propose a rigorous approach to mitigate these issues by introducing Faithful ViTs (FViTs). Briefly speaking, an FViT should have the following two properties: (1) The top-$k$ indices of its self-attention vector should remain mostly unchanged under input perturbation, indicating stable explanations; (2) The prediction distribution should be robust to perturbations. To achieve this, we propose a new method called Denoised Diffusion Smoothing (DDS), which adopts randomized smoothing and diffusion-based denoising. We theoretically prove that processing ViTs directly with DDS can turn them into FViTs. We also show that Gaussian noise is nearly optimal for both $\ell_2$ and $\ell_\infty$-norm cases. Finally, we demonstrate the effectiveness of our approach through comprehensive experiments and evaluations. Results show that FViTs are more robust against adversarial attacks while maintaining the explainability of attention, indicating higher faithfulness.
format Preprint
id arxiv_https___arxiv_org_abs_2311_17983
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Improving Interpretation Faithfulness for Vision Transformers
Hu, Lijie
Liu, Yixin
Liu, Ninghao
Huai, Mengdi
Sun, Lichao
Wang, Di
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Vision Transformers (ViTs) have achieved state-of-the-art performance for various vision tasks. One reason behind the success lies in their ability to provide plausible innate explanations for the behavior of neural architectures. However, ViTs suffer from issues with explanation faithfulness, as their focal points are fragile to adversarial attacks and can be easily changed with even slight perturbations on the input image. In this paper, we propose a rigorous approach to mitigate these issues by introducing Faithful ViTs (FViTs). Briefly speaking, an FViT should have the following two properties: (1) The top-$k$ indices of its self-attention vector should remain mostly unchanged under input perturbation, indicating stable explanations; (2) The prediction distribution should be robust to perturbations. To achieve this, we propose a new method called Denoised Diffusion Smoothing (DDS), which adopts randomized smoothing and diffusion-based denoising. We theoretically prove that processing ViTs directly with DDS can turn them into FViTs. We also show that Gaussian noise is nearly optimal for both $\ell_2$ and $\ell_\infty$-norm cases. Finally, we demonstrate the effectiveness of our approach through comprehensive experiments and evaluations. Results show that FViTs are more robust against adversarial attacks while maintaining the explainability of attention, indicating higher faithfulness.
title Improving Interpretation Faithfulness for Vision Transformers
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2311.17983