Skip to main content

Table 5 Sensitivity analysis with respect to the number of rounds r on the BREAST_CANCER dataset

From: Feature partitioning for robust tree ensembles and their certification in adversarial scenarios

b

Ak

Number of rounds r

  

1

15

30

f-FPF algorithm

1

0

0.956

0.939

0.939

 

1

0.842

0.921

0.921

 

2

0.000

0.675

0.719

 

3

0.000

0.114

0.158

2

0

0.93

0.939

0.939

 

1

0.904

0.930

0.930

 

2

0.754

0.842

0.842

 

3

0.123

0.404

0.474

3

0

0.947

0.939

0.939

 

1

0.912

0.921

0.921

 

2

0.833

0.86

0.86

 

3

0.675

0.772

0.763

4

0

0.956

0.939

0.939

 

1

0.895

0.904

0.904

 

2

0.825

0.868

0.868

 

3

0.728

0.807

0.807

5

0

0.930

0.939

0.930

 

1

0.912

0.895

0.895

 

2

0.842

0.860

0.868

 

3

0.746

0.816

0.816

h-FPF Algorithm

1

0

0.956

0.956

0.947

 

1

0.842

0.912

0.921

 

2

0.000

0.658

0.421

 

3

0.000

0.07

0.079

2

0

0.93

0.939

0.939

 

1

0.904

0.930

0.930

 

2

0.754

0.833

0.851

 

3

0.123

0.351

0.377

3

0

0.947

0.939

0.939

 

1

0.912

0.921

0.930

 

2

0.833

0.860

0.877

 

3

0.675

0.772

0.772

4

0

0.956

0.930

0.930

 

1

0.895

0.904

0.904

 

2

0.825

0.860

0.868

 

3

0.728

0.798

0.816

5

0

0.930

0.939

0.921

 

1

0.912

0.895

0.895

 

2

0.842

0.868

0.877

 

3

0.746

0.816

0.825