AI Paper Atlas

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

Countering the Communication Bottleneck in Federated Learning: A Highly Efficient Zero-Order Optimization Technique

Elissa Mhanna, Mohamad Assaad

JMLR 2024 · 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
use 2P-ZOFL (two scalars per round) instead of FedAvgfederated learning · FashionMNIST shirts vs sneakers (MLP, 197602 params)scalars uploaded per device: 59280600 -> 4000Improved
use 2P-ZOFL instead of FedZOfederated learning · FashionMNIST shirts vs sneakers (MLP, 197602 params)scalars uploaded per device: 148201500 -> 4000Improved
use 2P-ZOFL instead of FedAvg regarding rounds neededfederated learning · FashionMNIST shirts vs sneakers (MLP, 197602 params)communication rounds to convergence: 300 -> 2000Hurt
switch from IID to non-IID data distribution2P-ZOFL · FashionMNIST shirts vs sneakers (MLP, 197602 params)accuracy / convergence speedNo clear effect
increase wireless noise variance (ablation)2P-ZOFL · MNIST 0 vs 1 non-convex sigmoid, sigma_n^2 in {0.25, 1, 2.25, 3.1684, 10.0489}convergence speed: worst-case loss of 2.5% accuracyHurt
use constant instead of vanishing step sizes (ablation)2P-ZOFL · Mushroom, strictly convex logistic regressionconvergence speedImproved
add wireless noise (ablation)2P-ZOFL · Mushroom, strictly convex logistic regressionconvergence speedHurt
use constant step sizes with gradient-dominated loss (ablation)2P-ZOFL · MNIST 0 vs 1, kappa-gradient dominated lossconvergence rateImproved

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