Image Classification on CIFAR10 non-iid (Accuracy)
88.1AccuracyNo Attack
Evaluation results
| Method | Links | |
|---|---|---|
| No AttackAggregator=M-Krum, Architecture=ResNet-202024.04 | 88.1 | |
| Label FlipAggregator=M-Krum, Architecture=ResNet-202024.04 | 87.3 | |
| ROPAggregator=M-Krum, Architecture=ResNet-202024.04 | 87.3 | |
| No AttackAggregator=CC, Architecture=ResNet-202024.04 | 86.9 | |
| Min-MaxAggregator=M-Krum, Architecture=ResNet-202024.04 | 86.9 | |
| No AttackAggregator=TM, Architecture=ResNet-202024.04 | 86.7 | |
| No AttackAggregator=GAS (Krum), Architecture=ResNet-202024.04 | 86.7 | |
| No AttackAggregator=RFA, Architecture=ResNet-202024.04 | 86.4 | |
| No AttackAggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 86.3 | |
| No AttackAggregator=Bulyan, Architecture=ResNet-202024.04 | 86 | |
| Label FlipAggregator=GAS (Krum), Architecture=ResNet-202024.04 | 86 | |
| Min-MaxAggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 85.9 | |
| Label FlipAggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 85.5 | |
| Bit FlipAggregator=M-Krum, Architecture=ResNet-202024.04 | 85.2 | |
| Label FlipAggregator=CC, Architecture=ResNet-202024.04 | 85.2 | |
| Label FlipAggregator=RFA, Architecture=ResNet-202024.04 | 84.6 | |
| Bit FlipAggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 84.5 | |
| Bit FlipAggregator=GAS (Krum), Architecture=ResNet-202024.04 | 84.2 | |
| IPMAggregator=CC, Architecture=ResNet-202024.04 | 83.3 | |
| IPMAggregator=GAS (Krum), Architecture=ResNet-202024.04 | 83.3 | |
| IPMAggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 82.3 | |
| Min-MaxAggregator=Bulyan, Architecture=ResNet-202024.04 | 82.2 | |
| Bit FlipAggregator=CC, Architecture=ResNet-202024.04 | 81.8 | |
| Bit FlipAggregator=RFA, Architecture=ResNet-202024.04 | 80.9 | |
| Label FlipAggregator=Bulyan, Architecture=ResNet-202024.04 | 80.9 | |
| SLTBackbone=ResNet20, SFD/FedRolex constraint=1.02023.05 | 80.5 | |
| No AttackAggregator=CM, Architecture=ResNet-202024.04 | 79.9 | |
| Bit FlipAggregator=Bulyan, Architecture=ResNet-202024.04 | 79.8 | |
| IPMAggregator=TM, Architecture=ResNet-202024.04 | 77.9 | |
| Bit FlipAggregator=TM, Architecture=ResNet-202024.04 | 77 | |
| SLTBackbone=ResNet20, SFD/FedRolex constraint=0.52023.05 | 75.5 | |
| Small modelBackbone=ResNet20, SFD/FedRolex constraint=0.52023.05 | 73.6 | |
| Bit FlipAggregator=CM, Architecture=ResNet-202024.04 | 73.5 | |
| DHSADefense=LASA2024.04 | 73.4 | |
| IPMAggregator=M-Krum, Architecture=ResNet-202024.04 | 73.2 | |
| Label FlipAggregator=TM, Architecture=ResNet-202024.04 | 73 | |
| Label FlipAggregator=CM, Architecture=ResNet-202024.04 | 72.4 | |
| HSADefense=LASA2024.04 | 70.3 | |
| FEDADAVROptimiser=ADABELIEF, Server learning rate (ηs)=0.0012026.01 | 70.286 | |
| SLTBackbone=ResNet20, SFD/FedRolex constraint=0.252023.05 | 69.6 | |
| HO-SFLOptimization Order=Hybrid2026.03 | 69.6 | |
| FEDADAVROptimiser=ADAM, Server learning rate (ηs)=0.0012026.01 | 69.564 | |
| FEDADAVR-QUANTOptimiser=ADAM, Server learning rate (ηs)=0.0012026.01 | 69.564 | |
| SFLOptimization Order=First Order2026.03 | 69.3 | |
| FedAvgModel=AlexNet2025.12 | 69.22 | |
| FEDADAVR-QUANTOptimiser=LAMB, Server learning rate (ηs)=0.012026.01 | 69.175 | |
| FEDADAVR-QUANTOptimiser=ADAGRAD, Server learning rate (ηs)=0.0052026.01 | 68.293 | |
| FedOLFModel=AlexNet, TOA setting=no TOA2025.12 | 68.27 | |
| FEDADAVROptimiser=LAMB, Server learning rate (ηs)=0.012026.01 | 68.186 | |
| FEDADAVROptimiser=YOGI, Server learning rate (ηs)=0.0012026.01 | 68.121 | |
| FEDADAVROptimiser=LAMB, Server learning rate (ηs)=0.0052026.01 | 68.028 | |
| FEDADAVR-QUANTOptimiser=ADABELIEF, Server learning rate (ηs)=0.0012026.01 | 67.974 | |
| FEDADAVROptimiser=YOGI, Server learning rate (ηs)=0.0052026.01 | 67.969 | |
| FEDADAVR-QUANTOptimiser=YOGI, Server learning rate (ηs)=0.0012026.01 | 67.564 | |
| FEDADAVR-QUANTOptimiser=LAMB, Server learning rate (ηs)=0.0052026.01 | 67.456 | |
| FEDADAVROptimiser=ADAGRAD, Server learning rate (ηs)=0.0052026.01 | 67.409 | |
| FEDADAVROptimiser=ADABELIEF, Server learning rate (ηs)=0.0052026.01 | 67.349 | |
| FedOLFModel=AlexNet, TOA setting=TOA (s = 0.75)2025.12 | 66.6 | |
| IPMAggregator=RFA, Architecture=ResNet-202024.04 | 66.4 | |
| FEDADAVROptimiser=ADAM, Server learning rate (ηs)=0.0052026.01 | 65.622 | |
| FedVARPServer learning rate (ηs)=1.02026.01 | 65.367 | |
| FEDADAVR-QUANTOptimiser=ADAM, Server learning rate (ηs)=0.0052026.01 | 65.291 | |
| ROPAggregator=GAS (Krum), Architecture=ResNet-202024.04 | 64.6 | |
| FEDADAVR-QUANTOptimiser=ADABELIEF, Server learning rate (ηs)=0.0052026.01 | 64.362 | |
| HSA-Π_k(δ)Aggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 64.3 | |
| FEDADAVR-QUANTOptimiser=ADAM, Server learning rate (ηs)=0.012026.01 | 64.08 | |
| ALIEAggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 63.6 | |
| FEDADAVR-QUANTOptimiser=YOGI, Server learning rate (ηs)=0.0052026.01 | 63.411 | |
| FEDADAVR-QUANTOptimiser=YOGI, Server learning rate (ηs)=0.012026.01 | 63.33 | |
| FedOLFModel=AlexNet, TOA setting=TOA (s = 0.5)2025.12 | 63.12 | |
| Small modelBackbone=ResNet20, SFD/FedRolex constraint=0.252023.05 | 63.1 | |
| ROPAggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 62.5 | |
| IPMAggregator=CM, Architecture=ResNet-202024.04 | 61.9 | |
| CoCoFLModel=AlexNet2025.12 | 61.83 | |
| FEDADAVROptimiser=ADAM, Server learning rate (ηs)=0.012026.01 | 61.587 | |
| FEDADAVR-QUANTOptimiser=ADAGRAD, Server learning rate (ηs)=0.012026.01 | 61.392 | |
| FEDADAVROptimiser=ADAGRAD, Server learning rate (ηs)=0.012026.01 | 60.239 | |
| ALIEAggregator=CC, Architecture=ResNet-202024.04 | 59.7 | |
| FEDADAVR-QUANTOptimiser=ADABELIEF, Server learning rate (ηs)=0.012026.01 | 59.579 | |
| FEDADAVR-QUANTOptimiser=ADAGRAD, Server learning rate (ηs)=0.0012026.01 | 59.31 | |
| FEDADAVROptimiser=YOGI, Server learning rate (ηs)=0.012026.01 | 59.232 | |
| FEDADAVROptimiser=ADAGRAD, Server learning rate (ηs)=0.0012026.01 | 58.641 | |
| FEDADAVROptimiser=LAMB, Server learning rate (ηs)=0.0012026.01 | 58.506 | |
| FEDADAVR-QUANTOptimiser=LAMB, Server learning rate (ηs)=0.0012026.01 | 57.84 | |
| ALIEAggregator=GAS (Krum), Architecture=ResNet-202024.04 | 57 | |
| FEDADAVROptimiser=ADABELIEF, Server learning rate (ηs)=0.012026.01 | 55.696 | |
| ROPAggregator=TM, Architecture=ResNet-202024.04 | 55.6 | |
| HeteroFLModel=AlexNet2025.12 | 54.79 | |
| SLTBackbone=ResNet20, SFD/FedRolex constraint=0.1252023.05 | 52.4 | |
| FedVARPServer learning rate (ηs)=0.12026.01 | 51.629 | |
| ALIEAggregator=M-Krum, Architecture=ResNet-202024.04 | 49.7 | |
| DHSA-Π_k(δ, δ_FC^max)Aggregator=CM, Architecture=ResNet-202024.04 | 49 | |
| FedRolexBackbone=ResNet20, SFD/FedRolex constraint=0.52023.05 | 48.3 | |
| FjORDModel=AlexNet2025.12 | 46.3 | |
| IPMAggregator=Bulyan, Architecture=ResNet-202024.04 | 46.3 | |
| AdaptiveFLModel=AlexNet2025.12 | 45.82 | |
| Min-SumAggregator=GAS (Krum), Architecture=ResNet-202024.04 | 45.8 | |
| Min-SumAggregator=GAS (Bulyan), Architecture=ResNet-202024.04 | 45.1 | |
| Small modelBackbone=ResNet20, SFD/FedRolex constraint=0.1252023.05 | 44.7 | |
| ALIEAggregator=TM, Architecture=ResNet-202024.04 | 42.1 |