function labels=CKNN(PBags,NBags,testBags,Refs,Citers,indicator)
% CKNN Using the Citation KNN algorithm[1] to get the labels for bags in testBags, where minmum Hausdorff distance is used to measure the distances between bags
% CKNN takes,
% PBags - an Mx1 cell array where the jth instance of ith positive bag is stored in PBags{i}(j,:)
% NBags - an Nx1 cell array where the jth instance of ith negative bag is stored in NBags{i}(j,:)
% testBags - a Kx1 cell array where the jth instance of ith test bag is stored in testBags{i}(j,:)
% Refs - the number of referecnes for each test bag
% Citers - the number of citers for each test bag
% indicator - when the nearest neighbours have equal number of postive bags and negative bags, if indicator==1,then the output
% label is 1, otherwise 0, default=0
%
% and returns,
% labels - a Kx1 vector which correspond to the output label of each bag in testBags
%
% For more details, please reference to bibliography [1]
% [1] J. Wang and J.-D. Zucker. Solving the multiple-instance problem: a lazy learning approach. In: Proceedings of the 17th
% International Conference on Machine Learning, San Francisco, CA: Morgan Kaufmann, 1119-1125, 2000.
if(nargin<=4)
error('not enough input parameters');
end
if(nargin<=5)
indicator=0;
end
size1=size(PBags);
size2=size(NBags);
size3=size(testBags);
num_pbags=size1(1);
num_nbags=size2(1);
num_testbags=size3(1);
if((Refs>num_pbags+num_nbags)|(Citers>num_pbags+num_nbags))
erorr('too many Refs or Citers');
end
labels=zeros(num_testbags,1);
for i=1:num_testbags
num_bags=num_pbags+num_nbags+1;
dist=-eye(num_bags); %for every bag in testBAgs, initialize the distance matrix
for j=1:num_bags
if(j==1)
for k=(j+1):(num_pbags+1)
dist(j,k)=minHausdorff(testBags{i},PBags{k-1});
dist(k,j)=dist(j,k);
end
for k=(num_pbags+2):num_bags
dist(j,k)=minHausdorff(testBags{i},NBags{k-num_pbags-1});
dist(k,j)=dist(j,k);
end
else
if((j>=2)&(j<=num_pbags+1))
for k=(j+1):(num_pbags+1)
dist(j,k)=minHausdorff(PBags{j-1},PBags{k-1});
dist(k,j)=dist(j,k);
end
for k=(num_pbags+2):num_bags
dist(j,k)=minHausdorff(PBags{j-1},NBags{k-num_pbags-1});
dist(k,j)=dist(j,k);
end
else
for k=(j+1):num_bags
dist(j,k)=minHausdorff(NBags{j-num_pbags-1},NBags{k-num_pbags-1});
dist(k,j)=dist(j,k);
end
end
end
end
rp=0; %get the references and citers of current test bag
rn=0;
cp=0;
cn=0;
[temp,index]=sort(dist(1,:));
for ref=1:Refs
if(index(ref+1)<=num_pbags+1)
rp=rp+1;
else
rn=rn+1;
end
end
for pointer=2:(num_pbags+1)
[temp,index]=sort(dist(pointer,:));
if(find(index==1)<=Citers+1)
cp=cp+1;
end
end
for pointer=(num_pbags+2):num_bags
[temp,index]=sort(dist(pointer,:));
if(find(index==1)<=Citers+1)
cn=cn+1;
end
end
pos=rp+cp;
neg=rn+cn;
if(pos>neg)
labels(i,1)=1;
else
if(pos==neg)
if(indicator==1)
labels(i,1)=1;
end
end
end
end
