example_fcnn_training.cpp

 
#include <Pds/Ra>
#include <Pds/Nn>
    
int main(void)
{
    unsigned int M=8000;
    Pds::Vector Acc(M);
    std::string str;
    
    Pds::Matrix X(Pds::Ra::TextFormat,"../test/data_x_yinyang.txt");
    Pds::Matrix Y(Pds::Ra::TextFormat,"../test/data_y_yinyang.txt");
    Pds::Matrix Ys(Y.Size());
    
    std::vector<unsigned int> N={2,5,11,23,11,5,1};
    Pds::FCNn NN0=Pds::FCNn(N,Pds::Tanh,Pds::DTanh);
    
    Y=2*Y.Geq(0.5)-1.0;
    
    for(unsigned int k=0;k<M;k++)
    {
        NN0.Training(X,Y,0.1,0.01);
        
        bool res=NN0.Predict(X,Ys);
        if(res==false) std::cout<<"res==false"<<std::endl;
        
        Acc.Set(k,100.0*Pds::Accuracy(Ys,Y,0.1));
        
        // Progress bar
        str="acc: "+std::to_string(Acc.At(k));
        Pds::Ra::ProgressBarWithTime(32,k,M,false,str);
    }
    std::cout<<std::endl;
    
    NN0.Save("../test/NeuralNetwork1.txt"); 
    
    // Ploting
    Pds::Octave::XLabel="Iter";
    Pds::Octave::YLabel="Accuracy";
    Pds::Octave::Plot::CurveXY(Pds::LinSpace(0,M-1,M),Acc,"testando_acc.m","example_fcnn_training_Acc.png");
    
    Pds::Octave::XLabel="x_1";
    Pds::Octave::YLabel="x_2";
    Pds::Octave::Plot::PointsX2DY(X ,Y ,"testando_y.m","example_fcnn_training_Y.png");
    Pds::Octave::Plot::PointsX2DY(X,Ys,"testando_ys.m","example_fcnn_training_Ys.png");
    return 0;
}