We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Numerical Analysis

Title: A systematic approach to Lyapunov analyses of continuous-time models in convex optimization

Authors: Céline Moucer (SIERRA, ENPC), Adrien Taylor (SIERRA), Francis Bach (SIERRA)
Abstract: First-order methods are often analyzed via their continuous-time models, where their worst-case convergence properties are usually approached via Lyapunov functions. In this work, we provide a systematic and principled approach to find and verify Lyapunov functions for classes of ordinary and stochastic differential equations. More precisely, we extend the performance estimation framework, originally proposed by Drori and Teboulle [10], to continuous-time models. We retrieve convergence results comparable to those of discrete methods using fewer assumptions and convexity inequalities, and provide new results for stochastic accelerated gradient flows.
Subjects: Numerical Analysis (math.NA)
Cite as: arXiv:2205.12772 [math.NA]
  (or arXiv:2205.12772v1 [math.NA] for this version)

Submission history

From: Celine Moucer [view email]
[v1] Wed, 25 May 2022 13:40:29 GMT (65kb)

Link back to: arXiv, form interface, contact.