Table 11 Results for the Ablation analysis for the contribution of each component.

From: A personalized communication efficient federated learning framework with low rank adaptation for intelligent leukemia diagnosis

Model Variant

Accuracy (%)

F1-Score

AUC

MCC

Kappa

Communication Cost (MB)

Computation Time (min)

EfficientNet-B0

98.1

0.98

0.99

0.98

0.98

20.17

25

FedPer + EfficientNet-B0 (Without LoRA)

98.86

0.99

1

0.98

0.98

15.29

55

LoRA + EfficientNet-B0 (Without FedPer)

72.5

0.68

0.81

0.56

0.55

5.2

25

Proposed FedPer-B0 (FedPer + LoRA + EfficientNet-B0)

78.54

0.76

0.9

0.4

0.41

2.53

45

EfficientNet-B2

99.38

0.99

1

0.99

0.98

34.75

30

FedPer + EfficientNet-B2 (Without LoRA)

99.24

0.99

1

0.99

0.99

29.38

75

LoRA + EfficientNet-B2 (Without FedPer)

74.83

0.7

0.83

0.59

0.58

6.75

25

Proposed FedPer-B2 (FedPer + LoRA + EfficientNet-B2)

98.67

0.99

0.99

0.69

0.7

4.12

50