Algorithm 3.

Sample selection for kCPM

Input: Positive training samples 𝒟+, negative training samples 𝒟−, distance threshold r, order flag positive_first.
Output: Selected points set 𝒮
1:	𝒮 ←∅;
2:	if positive_first then
3:	𝒟 = 𝒟+ ∪𝒟−, where 𝒟+ occur first.
4:	else
5:	𝒟 = 𝒟− ∪𝒟+, where 𝒟− occur first.
6:	end if
7:	for all xi in 𝒟 in order do
8:	if minxj∈𝒮 d(xi, xj) > r then
9:	Add xi into 𝒮;
10:	end if
11:	end for