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. 2410.20659
29
1

A Statistical Analysis of Deep Federated Learning for Intrinsically Low-dimensional Data

28 October 2024
Saptarshi Chakraborty
Peter L. Bartlett
    FedML
ArXivPDFHTML
Abstract

Federated Learning (FL) has emerged as a groundbreaking paradigm in collaborative machine learning, emphasizing decentralized model training to address data privacy concerns. While significant progress has been made in optimizing federated learning, the exploration of generalization error, particularly in heterogeneous settings, has been limited, focusing mainly on parametric cases. This paper investigates the generalization properties of deep federated regression within a two-stage sampling model. Our findings highlight that the intrinsic dimension, defined by the entropic dimension, is crucial for determining convergence rates when appropriate network sizes are used. Specifically, if the true relationship between response and explanatory variables is charecterized by a β\betaβ-H\"older function and there are nnn independent and identically distributed (i.i.d.) samples from mmm participating clients, the error rate for participating clients scales at most as O~((mn)−2β/(2β+dˉ2β(λ)))\tilde{O}\left((mn)^{-2\beta/(2\beta + \bar{d}_{2\beta}(\lambda))}\right)O~((mn)−2β/(2β+dˉ2β​(λ))), and for non-participating clients, it scales as O~(Δ⋅m−2β/(2β+dˉ2β(λ))+(mn)−2β/(2β+dˉ2β(λ)))\tilde{O}\left(\Delta \cdot m^{-2\beta/(2\beta + \bar{d}_{2\beta}(\lambda))} + (mn)^{-2\beta/(2\beta + \bar{d}_{2\beta}(\lambda))}\right)O~(Δ⋅m−2β/(2β+dˉ2β​(λ))+(mn)−2β/(2β+dˉ2β​(λ))). Here, dˉ2β(λ)\bar{d}_{2\beta}(\lambda)dˉ2β​(λ) represents the 2β2\beta2β-entropic dimension of λ\lambdaλ, the marginal distribution of the explanatory variables, and Δ\DeltaΔ characterizes the dependence between the sampling stages. Our results explicitly account for the "closeness" of clients, demonstrating that the convergence rates of deep federated learners depend on intrinsic rather than nominal high-dimensionality.

View on arXiv
Comments on this paper