Non-ergodic linear convergence property of the delayed gradient descent under the strongly convexity and the Polyak-Łojasiewicz condition

DOI10.1142/S0219530524500106MaRDI QIDQ6587593zbMATH OpenFDO

Authors Hyung Jun Choi, Woocheol Choi, Jinmyoung Seok

Publication date 14 August 2024

Published in Analysis and Applications (Singapore) (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1142/S0219530524500106

zbMATH Keywords

convex programminggradient descent methodtime delay

Mathematics Subject Classification ID

Analysis of algorithms and problem complexity (68Q25) Convex programming (90C25)

Recommendations

• Conditions for linear convergence of the gradient method for non-convex optimization
• The error-feedback framework: SGD with delayed gradients
• Online strongly convex optimization with unknown delays
• Convergence rates for the heavy-ball continuous dynamics for non-convex optimization, under Polyak-Łojasiewicz condition
• Stochastic gradient descent with Polyak's learning rate

Cites work

• An Asynchronous Parallel Stochastic Coordinate Descent Algorithm
• Block coordinate type methods for optimization and learning
• Convex optimization: algorithms and complexity
• Decentralized consensus algorithm with delayed and stochastic gradients
• Decentralized Online Convex Optimization With Feedback Delays
• Distributed asynchronous incremental subgradient methods
• Event-triggered distributed online convex optimization with delayed bandit feedback
• Federated learning for minimizing nonsmooth convex loss functions
• Learning theory of randomized Kaczmarz algorithm
• Online Distributed Learning for Aggregative Games With Feedback Delays
• Online strongly convex optimization with unknown delays

This page was built for publication: Non-ergodic linear convergence property of the delayed gradient descent under the strongly convexity and the Polyak-Łojasiewicz condition

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6587593)