List your company
ResearchFormerly WizWand

Federated Image Classification on CIFAR100 (test)

70.78AccuracyFairGrad*
47.140853.277959.41565.5521Aug 16, 2025Oct 2, 2025Nov 18, 2025Jan 5, 2026Feb 21, 2026Apr 9, 2026May 27, 2026
Updated 4mo ago

Evaluation results

MethodLinks
FairGrad*Dirichlet concentration parameter (alpha)=0.1, Number of clients=1002025.0870.78———0.77
FairGradDirichlet concentration parameter (alpha)=0.5, Number of clients=1002025.0870.52———0.86
FairGrad*Dirichlet concentration parameter (alpha)=0.5, Number of clients=1002025.0869.67———0.78
FairGradDirichlet concentration parameter (alpha)=0.1, Number of clients=1002025.0869.38———0.91
FairGrad*Dirichlet concentration parameter (alpha)=0.05, Number of clients=1002025.0869.23———0.79
FairGradDirichlet concentration parameter (alpha)=0.05, Number of clients=1002025.0868.9———0.82
q-FFLDirichlet concentration parameter (alpha)=0.1, Number of clients=1002025.0867.7———0.86
FairLossDirichlet concentration parameter (alpha)=0.1, Number of clients=1002025.0867.55———0.88
AFLDirichlet concentration parameter (alpha)=0.5, Number of clients=1002025.0867.46———2.49
FairLoss*Dirichlet concentration parameter (alpha)=0.1, Number of clients=1002025.0867.41———0.9
FairLoss*Dirichlet concentration parameter (alpha)=0.5, Number of clients=1002025.0867.26———0.82
AAggFFDirichlet concentration parameter (alpha)=0.5, Number of clients=1002025.0867.22———0.89
q-FFLDirichlet concentration parameter (alpha)=0.5, Number of clients=1002025.0867.2———0.92
FairLoss*Dirichlet concentration parameter (alpha)=0.05, Number of clients=1002025.0867.16———0.88
FedAvgDirichlet concentration parameter (alpha)=0.05, Number of clients=1002025.0866.98———0.88
FairLossDirichlet concentration parameter (alpha)=0.5, Number of clients=1002025.0866.94———0.81
AAggFFDirichlet concentration parameter (alpha)=0.05, Number of clients=1002025.0866.83———0.93
FedAvgDirichlet concentration parameter (alpha)=0.1, Number of clients=1002025.0866.69———0.96
AAggFFDirichlet concentration parameter (alpha)=0.1, Number of clients=1002025.0866.68———0.85
FairLossDirichlet concentration parameter (alpha)=0.05, Number of clients=1002025.0866.46———0.81
q-FFLDirichlet concentration parameter (alpha)=0.05, Number of clients=1002025.0866.36———0.9
FedAvgDirichlet concentration parameter (alpha)=0.5, Number of clients=1002025.0866.35———0.92
AFLDirichlet concentration parameter (alpha)=0.1, Number of clients=1002025.0861.44———4.01
MCAAttack Name=LIE Attack, beta=0.2, b=0.12026.0558.81——0.3745—
Geometric MedianAttack Name=Gaussian Attack, beta=0.6, b=0.12026.0558.79——0.9003—
MCAAttack Name=Gaussian Attack, beta=0.6, b=0.12026.0558.76——0.4219—
PDR+MCAAttack Name=Gaussian Attack, beta=0.6, b=0.12026.0558.74——0.1611—
Geometric MedianAttack Name=LIE Attack, beta=0.6, b=0.12026.0558.73——0.8749—
PDR+Geometric MedianAttack Name=LIE Attack, beta=0.6, b=0.12026.0558.72——0.1616—
PDR+MCAAttack Name=LIE Attack, beta=0.6, b=0.12026.0558.69——0.144—
PDR+Geometric MedianAttack Name=Gaussian Attack, beta=0.6, b=0.12026.0558.67——0.1933—
MCAAttack Name=LIE Attack, beta=0.6, b=0.12026.0558.67——0.3671—
PDR+MCAAttack Name=Sign-flip Attack, beta=0.6, b=0.12026.0558.59——0.161—
PDR+MCAAttack Name=Gaussian Attack, beta=0.2, b=0.12026.0558.58——0.16—
PDR+Geometric MedianAttack Name=Gaussian Attack, beta=0.6, b=0.32026.0558.52——0.1637—
MCAAttack Name=Gaussian Attack, beta=0.2, b=0.12026.0558.52——0.4234—
PDR+MCAAttack Name=LIE Attack, beta=0.2, b=0.12026.0558.51——0.1495—
Geometric MedianAttack Name=LIE Attack, beta=0.2, b=0.12026.0558.5——0.9742—
PDR+Geometric MedianAttack Name=LIE Attack, beta=0.2, b=0.12026.0558.49——0.1595—
Geometric MedianAttack Name=Gaussian Attack, beta=0.6, b=0.32026.0558.45——0.918—
MCAAttack Name=FoE Attack, beta=0.6, b=0.12026.0558.41——0.4394—
PDR+Geometric MedianAttack Name=Gaussian Attack, beta=0.2, b=0.12026.0558.38——0.191—
MCAAttack Name=LIE Attack, beta=0.6, b=0.32026.0558.36——0.3669—
MCAAttack Name=Sign-flip Attack, beta=0.6, b=0.12026.0558.35——0.477—
BulyanAttack Name=Gaussian Attack, beta=0.6, b=0.12026.0558.31——8.5972—
Geometric MedianAttack Name=LIE Attack, beta=0.6, b=0.32026.0558.31——0.9026—
PDR+MCAAttack Name=Gaussian Attack, beta=0.6, b=0.32026.0558.31——0.137—
Geometric MedianAttack Name=Gaussian Attack, beta=0.2, b=0.12026.0558.3——0.9768—
MCAAttack Name=Gaussian Attack, beta=0.6, b=0.32026.0558.3——0.3709—
PDR+MCAAttack Name=FoE Attack, beta=0.6, b=0.12026.0558.3——0.1514—
PDR+MCAAttack Name=LIE Attack, beta=0.6, b=0.32026.0558.29——0.1356—
BulyanAttack Name=LIE Attack, beta=0.6, b=0.12026.0558.24——8.4861—
BulyanAttack Name=Sign-flip Attack, beta=0.2, b=0.12026.0558.22——8.2949—
BulyanAttack Name=Sign-flip Attack, beta=0.6, b=0.12026.0558.18——8.5215—
PDR+Geometric MedianAttack Name=LIE Attack, beta=0.6, b=0.32026.0558.04——0.164—
PDR+BulyanAttack Name=Gaussian Attack, beta=0.6, b=0.12026.0558.01——0.1216—
MCAAttack Name=FoE Attack, beta=0.2, b=0.12026.0557.96——0.443—
BulyanAttack Name=Gaussian Attack, beta=0.2, b=0.12026.0557.88——8.3849—
PDR+BulyanAttack Name=Sign-flip Attack, beta=0.6, b=0.12026.0557.87——0.115—
BulyanAttack Name=LIE Attack, beta=0.2, b=0.12026.0557.87——9.0898—
PDR+BulyanAttack Name=LIE Attack, beta=0.6, b=0.12026.0557.84——0.1104—
PDR+MCAAttack Name=FoE Attack, beta=0.2, b=0.12026.0557.82——0.1502—
PDR+MCAAttack Name=LIE Attack, beta=0.2, b=0.32026.0557.78——0.16—
Geometric MedianAttack Name=Gaussian Attack, beta=0.2, b=0.32026.0557.73——0.9659—
MCAAttack Name=Gaussian Attack, beta=0.2, b=0.32026.0557.69——0.3733—
PDR+Geometric MedianAttack Name=LIE Attack, beta=0.2, b=0.32026.0557.64——0.2033—
PDR+BulyanAttack Name=LIE Attack, beta=0.2, b=0.12026.0557.57——0.1104—
PDR+BulyanAttack Name=Gaussian Attack, beta=0.2, b=0.12026.0557.55——0.1216—
Geometric MedianAttack Name=LIE Attack, beta=0.2, b=0.32026.0557.53——0.9583—
PDR+BulyanAttack Name=Sign-flip Attack, beta=0.2, b=0.12026.0557.5——0.1158—
AFLDirichlet concentration parameter (alpha)=0.05, Number of clients=1002025.0857.5———4.94
PDR+Geometric MedianAttack Name=Gaussian Attack, beta=0.2, b=0.32026.0557.49——0.1678—
MCAAttack Name=Sign-flip Attack, beta=0.2, b=0.12026.0557.47——0.5364—
PDR+MCAAttack Name=Sign-flip Attack, beta=0.2, b=0.12026.0557.46——0.1806—
MCAAttack Name=LIE Attack, beta=0.2, b=0.32026.0557.45——0.3693—
PDR+MCAAttack Name=Gaussian Attack, beta=0.2, b=0.32026.0557.43——0.1465—
PDR+BulyanAttack Name=FoE Attack, beta=0.6, b=0.12026.0557.4——0.1143—
BulyanAttack Name=FoE Attack, beta=0.6, b=0.12026.0557.26——8.586—
BulyanAttack Name=FoE Attack, beta=0.2, b=0.12026.0557.18——8.3615—
PDR+MCAAttack Name=FoE Attack, beta=0.2, b=0.32026.0556.74——0.1463—
MCAAttack Name=FoE Attack, beta=0.6, b=0.32026.0556.69——0.439—
PDR+MCAAttack Name=FoE Attack, beta=0.6, b=0.32026.0556.5——0.1496—
MCAAttack Name=FoE Attack, beta=0.2, b=0.32026.0556.31——0.4352—
PDR+BulyanAttack Name=FoE Attack, beta=0.2, b=0.12026.0556——0.1164—
BulyanAttack Name=Sign-flip Attack, beta=0.6, b=0.32026.0554.74——8.3996—
BulyanAttack Name=Gaussian Attack, beta=0.6, b=0.32026.0554.34——8.4082—
BulyanAttack Name=LIE Attack, beta=0.6, b=0.32026.0554.13——8.3766—
PDR+Geometric MedianAttack Name=Sign-flip Attack, beta=0.6, b=0.12026.0553.8——0.1621—
Geometric MedianAttack Name=Sign-flip Attack, beta=0.6, b=0.12026.0553.39——0.9113—
PDR+Geometric MedianAttack Name=FoE Attack, beta=0.6, b=0.12026.0552.64——0.7713—
Geometric MedianAttack Name=FoE Attack, beta=0.6, b=0.12026.0552.5——3.4242—
PDR+BulyanAttack Name=Gaussian Attack, beta=0.6, b=0.32026.0551.93——0.1031—
BulyanAttack Name=Gaussian Attack, beta=0.2, b=0.32026.0551.09——8.1728—
PDR+BulyanAttack Name=Sign-flip Attack, beta=0.6, b=0.32026.0550.98——0.1031—
BulyanAttack Name=Sign-flip Attack, beta=0.2, b=0.32026.0550.87——8.1393—
PDR+Geometric MedianAttack Name=Sign-flip Attack, beta=0.2, b=0.12026.0550.75——0.1812—
Geometric MedianAttack Name=Sign-flip Attack, beta=0.2, b=0.12026.0550.74——0.9866—
PDR+BulyanAttack Name=LIE Attack, beta=0.6, b=0.32026.0550.7——0.1009—
BulyanAttack Name=LIE Attack, beta=0.2, b=0.32026.0550.04——8.1654—
Geometric MedianAttack Name=FoE Attack, beta=0.2, b=0.12026.0548.05——4.8993—