On the Role of Gaussian Covariates in Minimum Norm Interpolation
Event details
| Date | 16.05.2025 |
| Hour | 15:15 › 16:15 |
| Speaker | Gil Kur, ETH Zurich |
| Location | |
| Category | Conferences - Seminars |
| Event Language | English |
In the literature on benign overfitting in linear models, also referred to as minimum norm interpolation, it is typically assumed that the covariates follow a Gaussian distribution. Existing proofs heavily rely on the Gaussian Minimax Theorem (GMT), making them inapplicable to other distributions in the linear setting.
In our work, we are the first to establish matching rates for sub-Gaussian covariates in $\ell_p$-linear regression through a novel approach inspired by modern functional analysis.
In this talk, we provide an overview of this proof and explore the role of Gaussian covariates in benign overfitting from a purely geometric perspective.
Practical information
- Informed public
- Free
Organizer
- Yoav Zemel
Contact
- Maroussia Schaffner