Towards faster first order methods: A continuous-time model to interpolate between speed and function value restart
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914462126047232 |
|---|---|
| author | Maulén, Juan José Guo, Huiyuan Peypouquet, Juan |
| author_facet | Maulén, Juan José Guo, Huiyuan Peypouquet, Juan |
| contents | We introduce a new restarting scheme for a continuous inertial dynamics with Hessian driven-damping, and establish a linear convergence rate for the function values along the restarted trajectories. The proposed routine is implemented without knowing the strong convexity parameter, and is a generalization of existing speed restart schemes. It interpolates between speed and function value restarts, considerably delaying the restarting time, while preserving convergence and function value decrease. Numerical experiments show an improvement in the convergence rates for both continuous-time dynamical systems, and the associated accelerated first-order algorithms derived via time discretization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_11267 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Towards faster first order methods: A continuous-time model to interpolate between speed and function value restart Maulén, Juan José Guo, Huiyuan Peypouquet, Juan Optimization and Control We introduce a new restarting scheme for a continuous inertial dynamics with Hessian driven-damping, and establish a linear convergence rate for the function values along the restarted trajectories. The proposed routine is implemented without knowing the strong convexity parameter, and is a generalization of existing speed restart schemes. It interpolates between speed and function value restarts, considerably delaying the restarting time, while preserving convergence and function value decrease. Numerical experiments show an improvement in the convergence rates for both continuous-time dynamical systems, and the associated accelerated first-order algorithms derived via time discretization. |
| title | Towards faster first order methods: A continuous-time model to interpolate between speed and function value restart |
| topic | Optimization and Control |
| url | https://arxiv.org/abs/2506.11267 |