EURASIP Journal on Information Security

Table 6 Sensitivity analysis with respect to the training parameter b on the BREAST_CANCER dataset

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

\(\|{\mathcal {T}}\|\)	A_k	Parameter b
		1	2	3	4	5
f-FPF algorithm
50	0	0.939	0.939	0.947	0.939	0.956
	1	0.904	0.921	0.912	0.930	0.921
	2	0.737	0.851	0.868	0.868	0.860
	3	0.105	0.404	0.728	0.825	0.825
75	0	0.939	0.939	0.939	0.939	0.939
	1	0.912	0.930	0.912	0.904	0.904
	2	0.719	0.842	0.868	0.860	0.868
	3	0.167	0.404	0.754	0.816	0.816
100	0	0.939	0.939	0.939	0.939	0.939
	1	0.93	0.921	0.921	0.904	0.904
	2	0.465	0.842	0.860	0.886	0.868
	3	0.158	0.482	0.763	0.798	0.816
h-FPF Algorithm
50	0	0.947	0.939	0.947	0.947	0.947
	1	0.904	0.93	0.921	0.921	0.921
	2	0.711	0.842	0.868	0.86	0.877
	3	0.026	0.351	0.737	0.807	0.789
75	0	0.947	0.939	0.939	0.939	0.939
	1	0.912	0.93	0.921	0.904	0.904
	2	0.404	0.833	0.86	0.86	0.868
	3	0.096	0.351	0.746	0.798	0.789
100	0	0.947	0.939	0.939	0.947	0.930
	1	0.921	0.921	0.921	0.904	0.904
	2	0.412	0.842	0.868	0.877	0.868
	3	0.096	0.439	0.754	0.798	0.798

Back to article page