Table 2 Performance comparison of fed CL framework in cross school ideological and political Assessment.

From: A spatio-temporal graph diffusion and federated contrastive learning framework for cross-institutional educational evaluation

ε

τ

Aggregate weight α

Classification accuracy

Comparative Loss

(CL)

Model Consistency

(MC)

Privacy Leakage Risk

(PLR)

Adversarial Robustness

(AR)

1.5

0.7

0.18 ± 0.02

89.3 ± 1.2

1.23 ± 0.15

0.85 ± 0.03

0.12 ± 0.02

82.4 ± 1.5

2

0.5

0.22 ± 0.03

91.7 ± 0.9

0.95 ± 0.12

0.78 ± 0.04

0.18 ± 0.03

79.6 ± 1.8

1.2

0.8

0.15 ± 0.02

87.6 ± 1.5

1.45 ± 0.18

0.89 ± 0.02

0.09 ± 0.01

84.3 ± 1.2

1.8

0.6

0.20 ± 0.03

90.2 ± 1.1

1.07 ± 0.14

0.82 ± 0.03

0.15 ± 0.02

81.1 ± 1.6

1.5

0.7

0.18 ± 0.02

89.3 ± 1.2

1.23 ± 0.15

0.85 ± 0.03

0.12 ± 0.02

82.4 ± 1.5

Back to article page