Revisiting Generalization Power of a DNN in Terms of Symbolic Interactions
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910827701862400 |
|---|---|
| author | Cheng, Lei Zhang, Junpeng Ren, Qihan Zhang, Quanshi |
| author_facet | Cheng, Lei Zhang, Junpeng Ren, Qihan Zhang, Quanshi |
| contents | This paper aims to analyze the generalization power of deep neural networks (DNNs) from the perspective of interactions. Unlike previous analysis of a DNN's generalization power in a highdimensional feature space, we find that the generalization power of a DNN can be explained as the generalization power of the interactions. We found that the generalizable interactions follow a decay-shaped distribution, while non-generalizable interactions follow a spindle-shaped distribution. Furthermore, our theory can effectively disentangle these two types of interactions from a DNN. We have verified that our theory can well match real interactions in a DNN in experiments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_10162 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Revisiting Generalization Power of a DNN in Terms of Symbolic Interactions Cheng, Lei Zhang, Junpeng Ren, Qihan Zhang, Quanshi Machine Learning Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition This paper aims to analyze the generalization power of deep neural networks (DNNs) from the perspective of interactions. Unlike previous analysis of a DNN's generalization power in a highdimensional feature space, we find that the generalization power of a DNN can be explained as the generalization power of the interactions. We found that the generalizable interactions follow a decay-shaped distribution, while non-generalizable interactions follow a spindle-shaped distribution. Furthermore, our theory can effectively disentangle these two types of interactions from a DNN. We have verified that our theory can well match real interactions in a DNN in experiments. |
| title | Revisiting Generalization Power of a DNN in Terms of Symbolic Interactions |
| topic | Machine Learning Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2502.10162 |