Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2022
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2209.13517 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915863391633408 |
|---|---|
| author | Hirth, Johannes Hanika, Tom |
| author_facet | Hirth, Johannes Hanika, Tom |
| contents | We introduce \emph{conceptual views} as a formal framework grounded in Formal Concept Analysis for globally explaining neural networks. Experiments on twenty-four ImageNet models and Fruits-360 show that these views faithfully represent the original models, enable architecture comparison via Gromov--Wasserstein distance, and support abductive learning of human-comprehensible rules from neurons. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2209_13517 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | Conceptual Views of Neural Networks: A Framework for Neuro-Symbolic Analysis Hirth, Johannes Hanika, Tom Machine Learning Artificial Intelligence Computer Vision and Pattern Recognition 68T07 68T30 03G10 We introduce \emph{conceptual views} as a formal framework grounded in Formal Concept Analysis for globally explaining neural networks. Experiments on twenty-four ImageNet models and Fruits-360 show that these views faithfully represent the original models, enable architecture comparison via Gromov--Wasserstein distance, and support abductive learning of human-comprehensible rules from neurons. |
| title | Conceptual Views of Neural Networks: A Framework for Neuro-Symbolic Analysis |
| topic | Machine Learning Artificial Intelligence Computer Vision and Pattern Recognition 68T07 68T30 03G10 |
| url | https://arxiv.org/abs/2209.13517 |