Skip to main content

Table 8 Accuracy lower bound analysis on BREAST_CANCER. h-FPF with \(|\mathcal {T}| \approx 300\) and 8 leaves

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

 

k=1

k=2

k=3

b

\(\phantom {\dot {i}\!}ACC_{A_{1}}\)

\(\phantom {\dot {i}\!}ACC_{A_{1}}^{ELB}\)

\(\phantom {\dot {i}\!}ACC_{A_{1}}^{FLB}\)

\(\phantom {\dot {i}\!}ACC_{A_{2}}\)

\(\phantom {\dot {i}\!}ACC_{A_{2}}^{ELB}\)

\(\phantom {\dot {i}\!}ACC_{A_{2}}^{FLB}\)

\(\phantom {\dot {i}\!}ACC_{A_{3}}\)

\(\phantom {\dot {i}\!}ACC_{A_{3}}^{ELB}\)

\(\phantom {\dot {i}\!}ACC_{A_{3}}^{FLB}\)

1

0.912

0.895

0.868

0.746

0

0

0.175

0

0

2

0.904

0.904

0.904

0.842

0.833

0.816

0.632

0

0

3

0.912

0.912

0.904

0.868

0.851

0.842

0.789

0.737

0.701

4

0.904

0.904

0.895

0.860

0.860

0.860

0.807

0.798

0.781

5

0.895

0.895

0.895

0.868

0.868

0.860

0.816

0.807

0.798