ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2006.08950
17
171

Federated Accelerated Stochastic Gradient Descent

16 June 2020
Honglin Yuan
Tengyu Ma
    FedML
ArXivPDFHTML
Abstract

We propose Federated Accelerated Stochastic Gradient Descent (FedAc), a principled acceleration of Federated Averaging (FedAvg, also known as Local SGD) for distributed optimization. FedAc is the first provable acceleration of FedAvg that improves convergence speed and communication efficiency on various types of convex functions. For example, for strongly convex and smooth functions, when using MMM workers, the previous state-of-the-art FedAvg analysis can achieve a linear speedup in MMM if given MMM rounds of synchronization, whereas FedAc only requires M13M^{\frac{1}{3}}M31​ rounds. Moreover, we prove stronger guarantees for FedAc when the objectives are third-order smooth. Our technique is based on a potential-based perturbed iterate analysis, a novel stability analysis of generalized accelerated SGD, and a strategic tradeoff between acceleration and stability.

View on arXiv
Comments on this paper