Skip to main content

Table 19 Blended Blind FL results (constant weights)—Adult dataset

From: Blind Federated Learning without initial model

Agent

Size

% 1s

Accuracy

Accuracy

Precision

Precision

Pre-FL

Post-FL

Pre-FL

Post-FL

1

20%

75%

0.9166

0.9444

0.9285

0.9285

2

20%

69%

0.6944

0.9722

0.7333

0.9615

3

20%

69%

0.9444

0.9444

0.9259

0.8620

4

20%

86%

0.9444

0.9722

0.9677

1.0000

5

20%

83%

0.9722

0.9722

0.9677

0.9677

Avg.

0.8944

0.9611

0.9046

0.9439

1

13%

63%

0.5000

0.5217

1.0000

1.0000

2

30%

80%

0.3043

0.3478

1.0000

1.0000

3

21%

87%

0.3077

0.3205

1.0000

1.0000

4

11%

61%

0.5581

0.6047

0.6000

1.0000

5

24%

81%

0.3137

0.3529

0.8696

0.9005

Avg.

0.3968

0.4295

0.8939

0.9810

1

4%

75%

0.3125

0.3125

1.0000

0.0000

2

5%

68%

0.4210

0.4210

1.0000

0.6666

3

45%

77%

0.3374

0.3680

0.8214

0.9285

4

13%

51%

0.5217

0.5434

1.0000

1.0000

5

32%

85%

0.2991

0.3076

0.8695

0.8000

Avg.

0.3783

0.3905

0.9381

0.6790

1

1%

25%

0.3333

0.3333

0.0000

0.0000

2

5%

72%

0.3500

0.3500

1.0000

1.0000

3

45%

78%

0.3105

0.3354

0.8000

0.8461

4

14%

53%

0.5102

0.5102

0.6666

0.2500

5

35%

89%

0.1904

0.1904

1.0000

1.0000

Avg.

0.3389

0.3438

0.6933

0.6192

  1. Bold values indicate the best results