bicycle.cpp

#include <string>
#include <list>
#include <cstdio>
#include "maxent.h"

using namespace std;

void train_the_model_with_samples(ME_Model & model)
{
    ME_Sample s1("CAR");
    s1.add_feature("four wheels");              // binary feature
    s1.add_feature("blue");                     // binary feature
    s1.add_feature("length", 4.0);              // real-valued feature
    // This real-valued feature is almost equivalent to the following:
    //   s1.add_feature("length");
    //   s1.add_feature("length");
    //   s1.add_feature("length");
    //   s1.add_feature("length");

    ME_Sample s2("CAR");
    s2.add_feature("four wheels");
    s2.add_feature("red");
    s2.add_feature("length", 5.0);

    ME_Sample s3("BICYCLE");
    s3.add_feature("two wheels");
    s3.add_feature("red");
    s3.add_feature("length", 1.7);

    ME_Sample s4("BICYCLE");
    s4.add_feature("two wheels");
    s4.add_feature("yellow");
    s4.add_feature("length", 1.9);

    model.add_training_sample(s1);
    model.add_training_sample(s2);
    model.add_training_sample(s3);
    model.add_training_sample(s4);

    //model.use_l1_regularizer(1.0);

    model.train();
}

int main()
{
    ME_Model model;

    train_the_model_with_samples(model);

    // Classifying a new sample
    ME_Sample s;
    s.add_feature("two wheels");
    s.add_feature("blue");
    s.add_feature("length", 1.1);

    model.classify(s);
    cout << endl;
    cout << "it's a " << s.label << " !" << endl << endl;

    // You can get the probability distribution of a classification
    vector<double> vp = model.classify(s);
    for (int i = 0; i < model.num_classes(); i++) {
        cout << vp[i] << "\t" <<  model.get_class_label(i) << endl;
    }
    cout << endl;

    // You can save the model into a file.
    model.save_to_file("model");
    // You can, of course, load a model from a file.
    // Try replacing train_the_model_with_samples() with model.load_from_file("model").


    // If you want to see the weights of the features,
    list< pair< pair<string, string>, double > > fl;
    model.get_features(fl);
    for (list< pair< pair<string, string>, double> >::const_iterator i = fl.begin(); i != fl.end(); i++) {
        printf("%10.3f  %-10s %s\n", i->second, i->first.first.c_str(), i->first.second.c_str());
    }

    


}