AI Paper Atlas

Research map / Federated learning and privacy / Personalized and heterogeneous FL

Revisiting Weighted Aggregation in Federated Learning with Neural Networks

Zexi Li, Tao Lin, Xinyi Shang, Chao Wu

ICML 2023 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Non-IID federated learning · 18 citations (OpenAlex)

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 FedLAW instead of the strongest listed baselinefederated aggregation · FashionMNIST IID, MLPTop-1 test accuracy (%): 89.29 -> 88.51Hurt
Use FedLAW instead of the strongest listed baselinefederated aggregation · CIFAR-10 IID, ResNetTop-1 test accuracy (%): 77.94 -> 80.46Improved
Use FedLAW instead of the strongest listed baselinefederated aggregation · CIFAR-100 NonIID α=0.1, ResNetTop-1 test accuracy (%): 29.00 -> 25.31Hurt
Apply fixed global weight shrinking with γ=0.93 instead of γ=1.0 (ablation)FedAvg on CIFAR-10 IID α=100, SimpleCNN · CIFAR-10 IID α=100, E=2test accuracy: 65.53 -> 70.16Improved
Apply fixed global weight shrinking with γ=0.93 instead of γ=1.0 (ablation)FedAvg on CIFAR-10 NonIID α=1, SimpleCNN · CIFAR-10 NonIID α=1, E=2test accuracy: 65.58 -> 69.28Improved
Use a fixed shrinking factor that is too small or too large rather than the setting-specific best (ablation)global weight shrinking · CIFAR-10, varying IID and NonIID settingstest accuracyHurt
Add local weight decay to adaptive GWS (ablation)adaptive GWS on CIFAR-10 IID α=100 · CIFAR-10 IID α=100, E=2test accuracy: 71.47; 71.36; 71.35; 71.44; 71.54No clear effect
Add local weight decay to adaptive GWS (ablation)adaptive GWS on CIFAR-10 NonIID α=1 · CIFAR-10 NonIID α=1, E=2test accuracy: 70.31; 69.93; 70.44; 70.47; 69.99No clear effect
Combine adaptive global weight shrinking with local weight decay (ablation)FedAvg · CIFAR-10 IID and NonIID, E=2test accuracy: 66.43 -> 71.47; 65.35 -> 70.31Improved
Add SWA to FedLAW (ablation)FedLAW · FashionMNIST NonIID α=0.1, MLPTop-1 test accuracy (%): 86.30 -> 86.89Improved
Add SWA to FedLAW (ablation)FedLAW · CIFAR-10 NonIID α=0.1, ResNetTop-1 test accuracy (%): 52.83 -> 57.08Improved
Use early-stopped attentive LAW, then data-sized weights after the critical point (ablation)attentive LAW · CIFAR-10 hybrid client heterogeneitytest accuracyNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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