Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2405.14630 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913359817867264 |
|---|---|
| author | Karhadkar, Kedar Murray, Michael Montúfar, Guido |
| author_facet | Karhadkar, Kedar Murray, Michael Montúfar, Guido |
| contents | Bounds on the smallest eigenvalue of the neural tangent kernel (NTK) are a key ingredient in the analysis of neural network optimization and memorization. However, existing results require distributional assumptions on the data and are limited to a high-dimensional setting, where the input dimension $d_0$ scales at least logarithmically in the number of samples $n$. In this work we remove both of these requirements and instead provide bounds in terms of a measure of the collinearity of the data: notably these bounds hold with high probability even when $d_0$ is held constant versus $n$. We prove our results through a novel application of the hemisphere transform. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_14630 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Bounds for the smallest eigenvalue of the NTK for arbitrary spherical data of arbitrary dimension Karhadkar, Kedar Murray, Michael Montúfar, Guido Machine Learning Bounds on the smallest eigenvalue of the neural tangent kernel (NTK) are a key ingredient in the analysis of neural network optimization and memorization. However, existing results require distributional assumptions on the data and are limited to a high-dimensional setting, where the input dimension $d_0$ scales at least logarithmically in the number of samples $n$. In this work we remove both of these requirements and instead provide bounds in terms of a measure of the collinearity of the data: notably these bounds hold with high probability even when $d_0$ is held constant versus $n$. We prove our results through a novel application of the hemisphere transform. |
| title | Bounds for the smallest eigenvalue of the NTK for arbitrary spherical data of arbitrary dimension |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2405.14630 |