Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2503.00892 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913714663325696 |
|---|---|
| author | Costanza, Federico Simpson, Lachlan |
| author_facet | Costanza, Federico Simpson, Lachlan |
| contents | We introduce Riemannian Integrated Gradients (RIG); an extension of Integrated Gradients (IG) to Riemannian manifolds. We demonstrate that RIG restricts to IG when the Riemannian manifold is Euclidean space. We show that feature attribution can be phrased as an eigenvalue problem where attributions correspond to eigenvalues of a symmetric endomorphism. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_00892 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Riemannian Integrated Gradients: A Geometric View of Explainable AI Costanza, Federico Simpson, Lachlan Machine Learning Differential Geometry We introduce Riemannian Integrated Gradients (RIG); an extension of Integrated Gradients (IG) to Riemannian manifolds. We demonstrate that RIG restricts to IG when the Riemannian manifold is Euclidean space. We show that feature attribution can be phrased as an eigenvalue problem where attributions correspond to eigenvalues of a symmetric endomorphism. |
| title | Riemannian Integrated Gradients: A Geometric View of Explainable AI |
| topic | Machine Learning Differential Geometry |
| url | https://arxiv.org/abs/2503.00892 |