List your company
ResearchFormerly WizWand

Image Classification on FEMNIST (test)

97.31AccuracyQuPeD
41.82656.230570.63585.0395Jul 29, 2021Apr 28, 2022Jan 27, 2023Oct 27, 2023Jul 27, 2024Apr 26, 2025Jan 25, 2026
Updated 7mo ago

Evaluation results

MethodLinks
QuPeDPrecision=Full Precision (FP), Model=CNN12021.0797.31——
Per-FedAvgPrecision=Full Precision (FP), Model=CNN12021.0797.16——
QuPeDPrecision=2 Bits, Model=CNN12021.0796.73——
pFedMePrecision=Full Precision (FP), Model=CNN12021.0796.6——
Federated MLPrecision=Full Precision (FP), Model=CNN12021.0796.32——
QuPeLPrecision=2 Bits, Model=CNN12021.0796.1——
QuPeDPrecision=1 Bit, Model=CNN12021.0795.15——
QuPeDBackbone=CNN1, Quantization=FP2021.0794.93——
FedAvgPrecision=Full Precision (FP), Model=CNN12021.0794.92——
Local TrainingPrecision=Full Precision (FP), Model=CNN12021.0794.86——
QuPeDBackbone=CNN1, Quantization=2 Bits2021.0794.56——
QuPeLPrecision=1 Bit, Model=CNN12021.0794.06——
Local TrainingPrecision=2 Bits, Model=CNN12021.0793.95——
pFedMeBackbone=CNN1, Quantization=FP2021.0793.7——
Local TrainingPrecision=1 Bit, Model=CNN12021.0793——
Federated MLBackbone=CNN1, Quantization=FP2021.0792.73——
QuPeDBackbone=CNN1, Quantization=1 Bit2021.0792.52——
Per-FedAvgBackbone=CNN1, Quantization=FP2021.0792.1——
FedAvgBackbone=CNN1, Quantization=FP2021.0791.3——
FedEM2021.0879.9——
FedAvgPrivacy Budget=-, Backbone=CNN2025.0379.22—77.79
FedAvg2021.0878.6——
D-FedEMtopology=Binomial Erdős-Rényi graph (p=0.5), communication=fully decentralized2021.0877.2——
FedSVAPrivacy Budget=0.5, Backbone=CNN2025.0376.06—74.43
FedAvg+2021.0875.3——
FedSVAPrivacy Budget=0.4, Backbone=CNN2025.0375.3—73.88
pFedMe2021.0874.9——
KTA v2Comm. (MB)=94.62025.1174.5——
FedAvgComm. (MB)=154.32025.1174.3——
FedProxComm. (MB)=154.32025.1174.1——
FedSVAPrivacy Budget=0.3, Backbone=CNN2025.0373.79—73.18
Clustered FL2021.0873.5——
FedSVAPrivacy Budget=0.2, Backbone=CNN2025.0373.19—71.93
DPA DPPrivacy Budget=0.5, Backbone=CNN2025.0373.18—72.5
TV DPPrivacy Budget=0.5, Backbone=CNN2025.0372.63—72.39
DPA DPPrivacy Budget=0.3, Backbone=CNN2025.0372.25—71.75
DPA DPPrivacy Budget=0.4, Backbone=CNN2025.0372.25—71.92
Sens DPPrivacy Budget=0.5, Backbone=CNN2025.0372.21—72.05
TV DPPrivacy Budget=0.4, Backbone=CNN2025.0371.95—71.83
Sens DPPrivacy Budget=0.4, Backbone=CNN2025.0371.72—71.69
NbAFLPrivacy Budget=0.5, Backbone=CNN2025.0371.55—71.28
TV DPPrivacy Budget=0.3, Backbone=CNN2025.0371.35—69.22
NbAFLPrivacy Budget=0.4, Backbone=CNN2025.0371.33—70.85
Sens DPPrivacy Budget=0.3, Backbone=CNN2025.0371.22—71.03
Local2021.0871——
DPA DPPrivacy Budget=0.2, Backbone=CNN2025.0370.79—69.57
pFedBreD (mg)+AMModel Type=Personalized, Backbone=DNN, AM=true2022.1170.34——
pFedBreD (mg)+FTModel Type=Personalized, Backbone=DNN, Fine-tuning=true2022.1167.8——
Sens DPPrivacy Budget=0.2, Backbone=CNN2025.0367.33—68.92
FedAMPModel Type=Personalized, Backbone=DNN2022.1166.75——
NbAFLPrivacy Budget=0.3, Backbone=CNN2025.0366.16—62.49
Per-FedAvg+AMModel Type=Personalized, Backbone=DNN, AM=true2022.1166.03——
FedCCANumber of classes per client (N)=32026.0165.3——
pFedBreD (mg)Model Type=Personalized, Backbone=DNN2022.1165.29——
FEDSRNumber of classes per client (N)=32026.0165.1——
FEDMCNumber of classes per client (N)=32026.0164.8——
FEDREPNumber of classes per client (N)=32026.0164.6——
Per-FedAvg+FTModel Type=Personalized, Backbone=DNN, Fine-tuning=true2022.1163.99——
Per-FedAvgModel Type=Personalized, Backbone=DNN2022.1161.57——
FEDAMPNumber of classes per client (N)=32026.0160.7——
pFedBreD (mg)+AMModel Type=Personalized, Backbone=MCLR, AM=true2022.1160.64——
pFedMe+AMModel Type=Personalized, Backbone=DNN, AM=true2022.1160.46——
pFedBreD (mg)+FTModel Type=Global, Backbone=DNN, Fine-tuning=true2022.1160.16——
pFedBreD (mg)Model Type=Global, Backbone=DNN2022.1160.08——
FedAMPModel Type=Personalized, Backbone=MCLR2022.1160.04——
FedAvg+AMModel Type=Global, Backbone=DNN, AM=true2022.1160.03——
pFedBreD (mg)+FTModel Type=Personalized, Backbone=MCLR, Fine-tuning=true2022.1159.82——
pFedBreD (mg)+AMModel Type=Global, Backbone=DNN, AM=true2022.1159.53——
Per-FedAvg+AMModel Type=Global, Backbone=DNN, AM=true2022.1159.34——
Per-FedAvg+AMModel Type=Personalized, Backbone=MCLR, AM=true2022.1158.17——
FedAvgModel Type=Global, Backbone=DNN2022.1157.94——
pFedMe+AMModel Type=Global, Backbone=DNN, AM=true2022.1157.7——
Per-FedAvg+FTModel Type=Global, Backbone=DNN, Fine-tuning=true2022.1157.12——
pFedMe+AMModel Type=Personalized, Backbone=MCLR, AM=true2022.1156.91——
FEDDANNNumber of classes per client (N)=32026.0156.9——
pFedBreD (mg)Model Type=Personalized, Backbone=MCLR2022.1156.38——
Per-FedAvgModel Type=Global, Backbone=DNN2022.1156.3——
pFedBreD (mg)+AMModel Type=Global, Backbone=MCLR, AM=true2022.1156.23——
pFedMeModel Type=Global, Backbone=DNN2022.1156.2——
Per-FedAvg+AMModel Type=Global, Backbone=MCLR, AM=true2022.1156.16——
FedAvg+AMModel Type=Global, Backbone=MCLR, AM=true2022.1156.07——
pFedMe+FTModel Type=Global, Backbone=DNN, Fine-tuning=true2022.1155.89——
Per-FedAvg+FTModel Type=Personalized, Backbone=MCLR, Fine-tuning=true2022.1155.41——
pFedBreD (mg)Model Type=Global, Backbone=MCLR2022.1155.38——
FEDDANNumber of classes per client (N)=32026.0155.3——
pFedBreD (mg)+FTModel Type=Global, Backbone=MCLR, Fine-tuning=true2022.1155.1——
pFedMeModel Type=Personalized, Backbone=DNN2022.1154.72——
Per-FedAvgModel Type=Personalized, Backbone=MCLR2022.1154.34——
pFedMe+AMModel Type=Global, Backbone=MCLR, AM=true2022.1154.13——
FedAvgModel Type=Global, Backbone=MCLR2022.1153.96——
TV DPPrivacy Budget=0.2, Backbone=CNN2025.0353.88—50.92
pFedMe+FTModel Type=Personalized, Backbone=DNN, Fine-tuning=true2022.1153.79——
Per-FedAvg+FTModel Type=Global, Backbone=MCLR, Fine-tuning=true2022.1153.69——
Per-FedAvgModel Type=Global, Backbone=MCLR2022.1153.3——
pFedMe+FTModel Type=Personalized, Backbone=MCLR, Fine-tuning=true2022.1152.16——
pFedMeModel Type=Personalized, Backbone=MCLR2022.1151.7——
pFedMeModel Type=Global, Backbone=MCLR2022.1150.49——
pFedMe+FTModel Type=Global, Backbone=MCLR, Fine-tuning=true2022.1150.1——
LocalComm. (MB)=02025.1145.2——
FedEMModel Type=Global, Backbone=DNN2022.1143.96——