AI Paper Atlas

Research map / Federated learning and privacy / Communication-efficient FL

Decentralized Federated Learning With Distributed Aggregation Weight Optimization

Zhiyuan Zhai, Xiaojun Yuan, Xin Wang, Geoffrey Ye Li

TPAMI 2025 · Federated learning and privacy › Communication-efficient FL · approach: Local drift correction

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
increase the distance-decay exponent v from 2 to 10 (links more reliable)Proposed weight optimization · MNIST DFL, M=40, r=2optimized rho(W-bar): 0.54 -> 0.15Improved
use distributed subgradient optimization instead of centralized optimizationProposed weight optimization · FashionMNIST, M=200, r=4, v=2average and minimum test accuracyNo clear effect
optimize weights versus equal weights ignoring link reliability (ablation)DFL aggregation weights · FashionMNIST, M=200, r=4, v=2test accuracyImproved
optimize weights versus Metropolis-Hastings heuristic weights (ablation)DFL aggregation weights · FashionMNIST, M=200, r=4, v=2test accuracyImproved
lower rho(W-bar) of the aggregation matrix (ablation)DFL on MNIST · MNIST, M=40average and minimum test accuracyImproved
use dynamic (re-randomized positions) instead of static link reliabilityProposed weight optimization · MNIST, M=40, varying r and vminimum and average test accuracyImproved
increase link unreliability parameter rProposed weight optimization · MNIST, M=40convergence time of rho(W-bar)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy