Table 3:
Comparison of relative error (%) for federated learning (FL) and localized learning (LL) with respect to centralized learning (CL). The values denote average (standard deviation) over 10 iterations.
| Precision | Recall | Accuracy | |||||
| Dataset | Classifier | FL vs CL | LL vs CL | FL vs CL | LL vs CL | FL vs CL | LL vs CL |
| Opioid | SVM | 2.84(1.64) | 6.44(1.59) | 3.10(2.58) | 8.49(3.88) | 3.96(2.33) | 13.34(2.50) |
| Perceptron | 1.11(.73) | 7.16(2.49) | 7.32(5.45) | 9.06(6.06) | 5.31(3.99) | 12.88(4.58) | |
| LogReg | 1.86(1.11) | 11.28(2.56) | 3.21(2.81) | 11.55(3.37) | 2.66(2.54) | 12.55(4.49) | |
| Antipsychotic | SVM | 1.61(1.04) | 11.82(2.28) | 2.24(1.49) | 15.71(2.55) | 4.47(3.63) | 16.45(2.50) |
| Perceptron | 2.81(1.68) | 12.73(3.53) | 2.39(1.95) | 14.08(5.00) | 4.87(2.50) | 11.22(4.19) | |
| LogReg | 1.42(1.26) | 11.13(2.05) | 2.21(1.99) | 12.70(4.60) | 4.19(3.90) | 7.90(6.13) | |