Personalized Federated Learning with Semisupervised Distillation

Table 2

Test performance comparison of different thresholds for FedSD on MNIST and same model structures.

0.2	92.12	38.95
0.4	92.02	39.47
0.6	93.33	40.73
0.7	93.62	41.17
0.8	93.73	41.29
0.85	93.69	40.58
0.9	93.58	40.15
0.95	93.5	40.13