List your company
ResearchFormerly WizWand

Image Classification on CIFAR100 (Communication Efficiency)

67.9AccuracyFedMoSWA
34.142.87551.6560.425Jul 26, 2025Sep 1, 2025Oct 8, 2025Nov 14, 2025Dec 21, 2025Jan 27, 2026Mar 5, 2026
Updated 6mo ago

Evaluation results

MethodLinks
FedMoSWABackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0767.9———330
FedMoSWABackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0766.2———468
FedMoSWABackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0761.9———577
FedACGBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0761.7———518
MoFedSAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0760.1———603
FedSWABackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0759.8———574
FedACGBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0757.7———717
MoFedSAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0757.5———770
FedSWABackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0755.5———889
FedAvgBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0754.2———1,000
FedLESAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0753.3———1,000
QSGD[2]+FedBCGD+Backbone=LeNet-5, Heterogeneity (ρ)=0.6, Number of clients (N)=5, Total communication floats budget=200d2026.0353.1——56—
FedAvgBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0752.5———1,000
SCAFFOLDBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0752.3———1,000
QSGD[2]+FedBCGDBackbone=LeNet-5, Heterogeneity (ρ)=0.6, Number of clients (N)=5, Total communication floats budget=200d2026.0352.2——61—
FedACGBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0752.2———1,000
FedLESAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0752.1———1,000
FedSAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0751.9———1,000
MoFedSAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0751.5———1,000
SCAFFOLDBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0750.3———1,000
FedSWABackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0750.3———1,000
FedASAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0749.8———1,000
FedBCGD+Backbone=LeNet-5, Heterogeneity (ρ)=0.6, Number of clients (N)=5, Total communication floats budget=200d2026.0349.6——89—
FedSAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0749———1,000
FedBCGDBackbone=LeNet-5, Heterogeneity (ρ)=0.6, Number of clients (N)=5, Total communication floats budget=200d2026.0348.7——91—
FedLESAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0748.7———1,000
FedASAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0747.7———1,000
FedASAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0746.6———1,000
FedDynBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.6, Max Rounds=1000R, Target Accuracy=55%2025.0746.5———1,000
FedDynBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.3, Max Rounds=1000R, Target Accuracy=55%2025.0745.9———1,000
FedAvgBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0745.8———1,000
FedDynBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0745.8———1,000
SCAFFOLDBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0744.3———1,000
FedPAQBackbone=LeNet-5, Heterogeneity (ρ)=0.6, Number of clients (N)=5, Total communication floats budget=200d2026.0343.3——110—
TOP-kBackbone=LeNet-5, Heterogeneity (ρ)=0.6, Number of clients (N)=5, Total communication floats budget=200d2026.0342.2——112—
FedSAMBackbone=ResNet-18, Data Heterogeneity (Dirichlet coefficient)=0.1, Max Rounds=1000R, Target Accuracy=55%2025.0740.1———1,000
FedAvgBackbone=LeNet-5, Heterogeneity (ρ)=0.6, Number of clients (N)=5, Total communication floats budget=200d2026.0335.4————
FedAvgbeta=0.62022.03—81.67563.67——
FedDFbeta=0.62022.03—90445——
FedDynbeta=0.62022.03—56213.67——
FedFTGbeta=0.62022.03—55152.33——
FedGenbeta=0.62022.03—82571.33——
FedProxbeta=0.62022.03—81.67476——
MOONbeta=0.62022.03—83.67354——
SCAFFOLDbeta=0.62022.03—61.67186.33——