multigraph.m

function [result]=multigraph(X,s,alpha,beta,gamma)
% s is the true class label.
mv=size(X,1);
n=size(X{1},1);
Z=eye(n);
Zv=repmat(Z,[1,1,mv]);
c=length(unique(s));
wv=ones(mv,1)/mv;
%options = optimset( 'Algorithm','interior-point-convex','Display','off');
for ii=1:200
    Z(find(Z<0))=0;
    Z= (Z+Z')/2;
    Zold=Z;
    D = diag(sum(Z));
    L = D-Z;
    
    [F, temp, ev]=eig1(L, c, 0);
    for i=1:mv
        f=X{i};
        Zv(:,:,i)=(f*f'+alpha*eye(n)+beta*wv(i)*eye(n))\(beta*wv(i)*Z+f*f');
        T=Zv(:,:,i);
        T(find(T<0))=0;
        T=(T+T')/2;
        Zv(:,:,i)=T;
        wv(i)=1/2/norm(Zv(:,:,i)-Z,'fro');
        M(:,:,i)=wv(i)*Zv(:,:,i);
    end
    
    parfor ij=1:n
        all=distance(F,n,ij);      
        Z(:,ij)=(sum(M(:,ij,:),3)-gamma*all'/(4*beta))/sum(wv);
    end
    if ii>5 &((norm(Z-Zold,'fro')/norm(Zold,'fro'))<1e-3)
        break
    end
    
end
res=zeros(10,3);
for ij=1:10
actual_ids= kmeans(F, c, 'emptyaction', 'singleton', 'replicates', 1, 'display', 'off');
[res(ij,:)] = ClusteringMeasure( actual_ids,s);
end
result(1,1)=mean(res(:,1));result(1,2)=std(res(:,1));
result(2,1)=mean(res(:,2));result(2,2)=std(res(:,2));
result(3,1)=mean(res(:,3));result(3,2)=std(res(:,3));
end
function [all]=distance(F,n,ij);
  for ji=1:n
            all(ji)=(norm(F(ij,:)-F(ji,:)))^2;
  end
end