Manifold Integrated Gradients: Riemannian Geometry for Feature Attribution
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913352464203776 |
|---|---|
| author | Zaher, Eslam Trzaskowski, Maciej Nguyen, Quan Roosta, Fred |
| author_facet | Zaher, Eslam Trzaskowski, Maciej Nguyen, Quan Roosta, Fred |
| contents | In this paper, we dive into the reliability concerns of Integrated Gradients (IG), a prevalent feature attribution method for black-box deep learning models. We particularly address two predominant challenges associated with IG: the generation of noisy feature visualizations for vision models and the vulnerability to adversarial attributional attacks. Our approach involves an adaptation of path-based feature attribution, aligning the path of attribution more closely to the intrinsic geometry of the data manifold. Our experiments utilise deep generative models applied to several real-world image datasets. They demonstrate that IG along the geodesics conforms to the curved geometry of the Riemannian data manifold, generating more perceptually intuitive explanations and, subsequently, substantially increasing robustness to targeted attributional attacks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_09800 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Manifold Integrated Gradients: Riemannian Geometry for Feature Attribution Zaher, Eslam Trzaskowski, Maciej Nguyen, Quan Roosta, Fred Machine Learning Human-Computer Interaction Differential Geometry In this paper, we dive into the reliability concerns of Integrated Gradients (IG), a prevalent feature attribution method for black-box deep learning models. We particularly address two predominant challenges associated with IG: the generation of noisy feature visualizations for vision models and the vulnerability to adversarial attributional attacks. Our approach involves an adaptation of path-based feature attribution, aligning the path of attribution more closely to the intrinsic geometry of the data manifold. Our experiments utilise deep generative models applied to several real-world image datasets. They demonstrate that IG along the geodesics conforms to the curved geometry of the Riemannian data manifold, generating more perceptually intuitive explanations and, subsequently, substantially increasing robustness to targeted attributional attacks. |
| title | Manifold Integrated Gradients: Riemannian Geometry for Feature Attribution |
| topic | Machine Learning Human-Computer Interaction Differential Geometry |
| url | https://arxiv.org/abs/2405.09800 |