Power of Generalized Smoothness in Stochastic Convex Optimization: First- and Zero-Order Algorithms
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912388789305344 |
|---|---|
| author | Lobanov, Aleksandr Gasnikov, Alexander |
| author_facet | Lobanov, Aleksandr Gasnikov, Alexander |
| contents | This paper is devoted to the study of stochastic optimization problems under the generalized smoothness assumption. By considering the unbiased gradient oracle in Stochastic Gradient Descent, we provide strategies to achieve in bounds the summands describing linear rate. In particular, in the case $L_0 = 0$, we obtain in the convex setup the iteration complexity: $N = \mathcal{O}\left(L_1R \log\frac{1}{\varepsilon} + \frac{L_1 c R^2}{\varepsilon}\right)$ for Clipped Stochastic Gradient Descent and $N = \mathcal{O}\left(L_1R \log\frac{1}{\varepsilon}\right)$ for Normalized Stochastic Gradient Descent. Furthermore, we generalize the convergence results to the case with a biased gradient oracle, and show that the power of $(L_0,L_1)$-smoothness extends to zero-order algorithms. Finally, we demonstrate the possibility of linear convergence in the convex setup through numerical experimentation, which has aroused some interest in the machine learning community. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_18198 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Power of Generalized Smoothness in Stochastic Convex Optimization: First- and Zero-Order Algorithms Lobanov, Aleksandr Gasnikov, Alexander Optimization and Control This paper is devoted to the study of stochastic optimization problems under the generalized smoothness assumption. By considering the unbiased gradient oracle in Stochastic Gradient Descent, we provide strategies to achieve in bounds the summands describing linear rate. In particular, in the case $L_0 = 0$, we obtain in the convex setup the iteration complexity: $N = \mathcal{O}\left(L_1R \log\frac{1}{\varepsilon} + \frac{L_1 c R^2}{\varepsilon}\right)$ for Clipped Stochastic Gradient Descent and $N = \mathcal{O}\left(L_1R \log\frac{1}{\varepsilon}\right)$ for Normalized Stochastic Gradient Descent. Furthermore, we generalize the convergence results to the case with a biased gradient oracle, and show that the power of $(L_0,L_1)$-smoothness extends to zero-order algorithms. Finally, we demonstrate the possibility of linear convergence in the convex setup through numerical experimentation, which has aroused some interest in the machine learning community. |
| title | Power of Generalized Smoothness in Stochastic Convex Optimization: First- and Zero-Order Algorithms |
| topic | Optimization and Control |
| url | https://arxiv.org/abs/2501.18198 |