Table 6 Average results for scenario E on all five datasets over 100 aggregation rounds
From: FedOcw: optimized federated learning for cross-lingual speech-based Parkinson’s disease detection
Scenario E | Local | Federated learning methods | Centralized learning | ||||
|---|---|---|---|---|---|---|---|
FedAvg | FedProx | Scaffold | FedNova | FedOcw | |||
Accuracy | 65.68 (61.07–70.58) | 69.58 (57.41–73.12) | 69.35 (60.35–73.85) | 67.37 (59.34–70.72) | 46.09 (43.04–75.64) | 72.63 (62.5–75.58 | 68.31 (67.76–68.87) |
F1-score | 61.55 (56.46–67.84) | 63.15 (37.09–68.61) | 62.72 (43.83–69.33) | 62.39 (41.24–66.82) | 33.9 (29.87–72.09) | 68.16 (48.74–72.46) | 66.83 (65.53–67.6) |
Specifity | 65.2 (60.8–70.24) | 65.4 (50.47–69.41) | 65.05 (53.08–70.36) | 64.59 (52.71–68.31) | 52.14 (50–73.73) | 69.41 (56.37–73.4) | 67 (66.05–67.68) |
Sensitivity | 66.06 (61.3–72.22) | 70.3 (36.3–76.88) | 67.81 (52.04–75.81) | 66.64 (47.51–72.15) | 26.96 (21.52–77.27) | 75.1 (55.89–79.25) | 68.79 (68.3–69.42) |
Mcc | 0.312 (0.221–0.422) | 0.357 (0.026–0.433) | 0.339 (0.093–0.45) | 0.325 (0.01–0.392) | 0.045 (0–0.504) | 0.435 (0.173–0.508 | 0.357 (0.348–0.369) |
