List your company
ResearchFormerly WizWand

Image Classification on MNIST (test) (Performance by Client Class Count)

99.05Acc (3 Classes/Client)QuPeD
98.301298.495698.6998.8844Jul 29, 2021
Updated 8mo ago

Evaluation results

MethodLinks
QuPeDPrecision=FP, Backbone=CNN12021.0799.0598.89
Federated MLPrecision=FP, Backbone=CNN12021.079998.84
pFedMePrecision=FP, Backbone=CNN12021.0798.9898.82
QuPeDPrecision=2 Bits, Backbone=CNN12021.0798.9698.67
QuPeLPrecision=2 Bits, Backbone=CNN12021.0798.9598.61
Per-FedAvgPrecision=FP, Backbone=CNN12021.0798.8298.93
Local TrainingPrecision=FP, Backbone=CNN12021.0798.7998.66
FedAvgPrecision=FP, Backbone=CNN12021.0798.6498.65
QuPeDPrecision=1 Bit, Backbone=CNN12021.0798.5798.25
Local TrainingPrecision=2 Bits, Backbone=CNN12021.0798.5398.37
Local TrainingPrecision=1 Bit, Backbone=CNN12021.0798.4197.95
QuPeLPrecision=1 Bit, Backbone=CNN12021.0798.3398.11