Implicit Differentiation for Hyperparameter Tuning the Weighted Graphical Lasso
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2023
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866929383811317760 |
|---|---|
| author | Pouliquen, Can Gonçalves, Paulo Massias, Mathurin Vayer, Titouan |
| author_facet | Pouliquen, Can Gonçalves, Paulo Massias, Mathurin Vayer, Titouan |
| contents | We provide a framework and algorithm for tuning the hyperparameters of the Graphical Lasso via a bilevel optimization problem solved with a first-order method. In particular, we derive the Jacobian of the Graphical Lasso solution with respect to its regularization hyperparameters. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2307_02130 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Implicit Differentiation for Hyperparameter Tuning the Weighted Graphical Lasso Pouliquen, Can Gonçalves, Paulo Massias, Mathurin Vayer, Titouan Machine Learning Optimization and Control We provide a framework and algorithm for tuning the hyperparameters of the Graphical Lasso via a bilevel optimization problem solved with a first-order method. In particular, we derive the Jacobian of the Graphical Lasso solution with respect to its regularization hyperparameters. |
| title | Implicit Differentiation for Hyperparameter Tuning the Weighted Graphical Lasso |
| topic | Machine Learning Optimization and Control |
| url | https://arxiv.org/abs/2307.02130 |