Boundary between noise and information applied to filtering neural network weight matrices
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2022
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914645438103552 |
|---|---|
| author | Staats, Max Thamm, Matthias Rosenow, Bernd |
| author_facet | Staats, Max Thamm, Matthias Rosenow, Bernd |
| contents | Deep neural networks have been successfully applied to a broad range of problems where overparametrization yields weight matrices which are partially random. A comparison of weight matrix singular vectors to the Porter-Thomas distribution suggests that there is a boundary between randomness and learned information in the singular value spectrum. Inspired by this finding, we introduce an algorithm for noise filtering, which both removes small singular values and reduces the magnitude of large singular values to counteract the effect of level repulsion between the noise and the information part of the spectrum. For networks trained in the presence of label noise, we indeed find that the generalization performance improves significantly due to noise filtering. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2206_03927 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | Boundary between noise and information applied to filtering neural network weight matrices Staats, Max Thamm, Matthias Rosenow, Bernd Disordered Systems and Neural Networks Machine Learning Deep neural networks have been successfully applied to a broad range of problems where overparametrization yields weight matrices which are partially random. A comparison of weight matrix singular vectors to the Porter-Thomas distribution suggests that there is a boundary between randomness and learned information in the singular value spectrum. Inspired by this finding, we introduce an algorithm for noise filtering, which both removes small singular values and reduces the magnitude of large singular values to counteract the effect of level repulsion between the noise and the information part of the spectrum. For networks trained in the presence of label noise, we indeed find that the generalization performance improves significantly due to noise filtering. |
| title | Boundary between noise and information applied to filtering neural network weight matrices |
| topic | Disordered Systems and Neural Networks Machine Learning |
| url | https://arxiv.org/abs/2206.03927 |