CV_Lasso_logistic.m

function [ opt_s,Mse ] = CV_Lasso_logistic(X,y,Lambda,beta_path)

%%%%%%%%%%%%%%     K cross validation    %%%%%%%%%%%%%%%%%%%%%%%%%%%%
k=10;            %%% K-fold%%%
[n,p] = size(X);
valida_n=floor(n/k);
sample_sequence=1:n;

for j=1:length(Lambda)
    lambda=Lambda(j);
    beta_ini=beta_path(:,j);
    for i=1:k
      if i<=k-1
          validation_seq=sample_sequence(:,(i-1)*valida_n+1:i*valida_n);
      else
          validation_seq=sample_sequence(:,(i-1)*valida_n+1:n);
      end
      train_seq=setdiff(sample_sequence,validation_seq);
      X_train = X(train_seq,:);
      y_train = y(train_seq);
      X_validation= X(validation_seq, :);
      y_validation = y(validation_seq);
      [b,intercept]=Lasso_CD_logistic(X_train,y_train,lambda,beta_ini);
      l=intercept+X_validation*b;
      prob=exp(l)./(1 + exp(l));
      for m=1:size(y_validation,1)
        if prob(m)>0.5
            test_y(m)=1;
        else
            test_y(m)=0;
        end
      end
      error=test_y'-y_validation;
      Mse(i,j)=sum(abs(error));       %Mse(i,j)=sum(abs(prob-y_validation));
      test_y=0;
    end
end
[d,opt_s]=min(sum(Mse,1));
Mse=sum(Mse,1);
end