Mathematics > Optimization and Control
[Submitted on 28 Nov 2024 (v1), last revised 2 Mar 2025 (this version, v2)]
Title:Advances in Nonmonotone Proximal Gradient Methods merely with Local Lipschitz Assumptions in the Presense of Kurdyka-Łojasiewicz Property: A Study of Average and Max Line Search
View PDF HTML (experimental)Abstract:The proximal gradient method is a standard approach to solve the composite minimization problems where the objective function is the sum of a continuously differentiable function and a lower semicontinuous, extended-valued function. For both monotone and nonmonotone proximal gradient methods, the convergence theory has traditionally replied heavily on the assumption of global Lipschitz continuity. Recent works have shown that the monotone proximal gradient method, even when the local Lipschitz continuity (rather than global) is assumed, converges to the stationarity globally in the presence of Kurdyka-Łojasiewicz Property. However, how to extend these results from monotone proximal gradient method to nonmonotone proximal gradient method (NPG) remains an open question. In this manuscript, we consider two types of NPG: those combined with average line search and max line search, respectively. By partitioning of indices into two subsets, one of them aims to achieve a decrease in the functional sequence, we establish the global convergence and rate-of-convergence (same as the monotone version) results under the KL property, merely requiring the local Lipschitz assumption, and without an a priori knowledge of the iterative sequence being bounded. When our work is almost done, we noticed that [17] presented the analogous results for the NPG with average line search, whose partitioning of index set is totally different with ours. Drawing upon the findings in this manuscript and [17], we confidently conclude that the convergence theory of NPG is independent on the specific partitioning of the index set.
Submission history
From: Xiaoxi Jia [view email][v1] Thu, 28 Nov 2024 16:44:27 UTC (35 KB)
[v2] Sun, 2 Mar 2025 22:12:57 UTC (36 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.