/**
* テストの間の処理
*/
void Test_mainloop()
{
term_keymode(TRUE);
Test_begin();
if (!test.intprt.status) return;
Test_typing();
if (!test.intprt.status) return;
Test_notify_end();
if (!test.intprt.status) return;
check();
Test_result();
Test_error_check();
if (!test.intprt.status) return;
}
int main(int argc, char *argv[])
{
cout<<"training1 file,training2 file,testing file, number of training1, number of training2,number of testing"<<endl;
cout<<" number of feature, iteration,rho1,rho2,esp"<<endl;
string nametrain=argv[1];
string nametrain_T=argv[2];
string nametest=argv[3];
int num_train_rows=atoi(argv[4]); // Total number of train instances in the dataset.
int num_train_rows2=atoi(argv[5]); // Total number of train instances in the dataset.
int test_rows=atoi(argv[6]); // Total number of test instances in the dataset.
int num_cols=atoi(argv[7]); // Total number of features in the dataset.
int max_iter=atoi(argv[8]);//Variable for maximum number of iteration
double weight=atof(argv[9]); // step size (0.2)parameter.
double rho2=atof(argv[10]);
double eps=atof(argv[11]);
int num_lambda=atoi(argv[12]);
double step=atof(argv[13]);
ColumnVector Betainit(num_cols);
Betainit.setZero(num_cols);
ColumnVector Betainit_T(num_cols);
Betainit_T.setZero(num_cols);
ColumnVector Betatemp_T(num_cols);
Betatemp_T.setZero(num_cols);
Cox21 newcox(nametrain,nametrain_T,nametest,num_train_rows,num_train_rows2,num_cols);
Coxnotied test_part;
test_part.create_from_file(nametest,test_rows,num_cols);
time_t t1,t2;
MatrixXd W(num_cols,2);
MatrixXd Topfeatures(52,num_lambda);
MatrixXd allrun(4,num_lambda);
MatrixXd runinfor(56,num_lambda);
MatrixXd Test_result(test_rows,(num_lambda+2));
MatrixXd Train_T_result(num_train_rows2,(num_lambda+2));
MatrixXd ALL_Beta(num_cols,num_lambda);
Test_result.col(0)=test_part.Times;
Test_result.col(1)=test_part.Status;
Train_T_result.col(0)=newcox.train_part_T.Times;
Train_T_result.col(1)=newcox.train_part_T.Status;
double lambda_max=newcox.find_lambda_max(rho2,weight);
double cindex=0.0;
double lambda=0.0;
double lambda_old=0.0;
t1=time(0);
for(int i=0;i<(num_lambda);i++){
lambda_old=lambda_max*pow (step, double(i)/double(num_lambda));
lambda=lambda_max*pow (step, double(i+1)/double(num_lambda));
W=newcox.cox_L21_strong (Betainit, Betainit_T, rho2,lambda, lambda_old,max_iter,eps,weight);
Betainit=W.col(0);
Betainit_T=W.col(1);
ALL_Beta.col(i)=W.col(1);
allrun(0,i)=lambda;
allrun(1,i)=newcox.obj;
allrun(2,i)=newcox.funcv2;
allrun(3,i)=newcox.featureleft;
cout<<i<<endl;
}
t2=time(0);
float diff((float)t2-(float)t1);
cout<<"running time is"<<diff<<endl;
for(int i=0;i<num_lambda;i++){
Betatemp_T=ALL_Beta.col(i);
// record the estimated value for BS in R
test_part.estimate(Betatemp_T);
Test_result.col((i+2))=test_part.predict;
newcox.train_part_T.estimate(Betatemp_T);
Train_T_result.col((i+2))=newcox.train_part_T.predict;
// recorde the cindex, objective, nonzero, and top 50 features.
cindex=test_part.GetCindex_yan();
Topfeatures.col(i)=find_top_n(Betatemp_T, 52, cindex);
//double nonzero=Betainit_T.nonZeros();
}
runinfor<<allrun,
Topfeatures;
//****************************************************************************************************************************//
// Prepare the output file at here
string name = argv[1];
string out=boost::lexical_cast<string>(step);
string outname1=name+out+"_train_result_new.txt";
string outname2=name+out+"_test_result_new.txt";
string outname3=name+out+"_record_new.txt";
string outname4=name+out+"_runtime.txt";
const char *outputname1=outname1.c_str();
const char *outputname2=outname2.c_str();
const char *outputname3=outname3.c_str();
const char *outputname4=outname4.c_str();
ofstream outfile1,outfile2,outfile3,outfile4;
outfile1.open(outputname1);
outfile2.open(outputname2);
outfile3.open(outputname3);
outfile4.open(outputname4);
outfile1<<Train_T_result<<endl;
outfile2<<Test_result<<endl;
outfile3<<runinfor<<endl;
outfile4<<"running time is "<<diff<<endl;
//cout<<runinfor<<endl;
outfile1.close();
outfile2.close();
outfile3.close();
outfile4.close();
return 0;
}