#include "opencv2/core/core.hpp"

#include "opencv2/ml/ml.hpp"

using namespace cv;

int main()

{

        Mat_<float> trainingSet1( Size( 2, 25 ) );

        Mat_<int> response1( Size( 1, 25 ) );

        Mat_<float> trainingSet2( Size( 2, 25 ) );

        Mat_<int> response2( Size( 1, 25 ) );

        Mat_<float> sample( Size( 2, 1000 ) );

        RNG rng = RNG( (int) time(0) );

        //prepare data

        for( size_t i = 0; i < trainingSet1.rows; i++ )

        {

                for( size_t j = 0; j < trainingSet1.cols; j++ )

                {

                        float x = rng.uniform( 0, 1000 );

                        trainingSet1( i, j ) = x;

                }

        }

        for( size_t i = 0; i < trainingSet2.rows; i++ )

        {

                for( size_t j = 0; j < trainingSet2.cols; j++ )

                {

                        float x = rng.uniform( 0, 1000 );

                        trainingSet2( i, j ) = x;

                }

        }

        for( size_t i = 0; i < response1.rows; i++ )

        {

                int x = rng.uniform( 0, 2 );

                response1( i, 1 ) = x;

        }

        for( size_t i = 0; i < response2.rows; i++ )

        {

                int x = rng.uniform( 0, 2 );

                response2( i, 1 ) = x;

        }

        //end

        Mat var_types1( 1, trainingSet1.cols + 1, CV_8UC1, Scalar( CV_VAR_ORDERED ) );

        var_types1.at<uchar>( trainingSet1.cols ) = CV_VAR_CATEGORICAL;

        Mat var_types2( 1, trainingSet2.cols + 1, CV_8UC1, Scalar( CV_VAR_ORDERED ) );

        var_types2.at<uchar>( trainingSet2.cols ) = CV_VAR_CATEGORICAL;

        CvBoostParams  params( CvBoost::REAL, // boost_type

                           25, // weak_count

                           0.8, // weight_trim_rate

                           1, // max_depth

                           false, //use_surrogates

                           0 // priors

                         );

        CvBoost boostModel;

        //first train

        boostModel.train( trainingSet1, CV_ROW_SAMPLE, response1, Mat(), Mat(), var_types1, Mat(), params, false );

        //predict

        for( size_t i = 0; i < sample.rows; i++ )

        {

                boostModel.predict( sample.row(i), Mat(), Range::all(), false, true );

        }

        //second train

        boostModel.train( trainingSet2, CV_ROW_SAMPLE, response2, Mat(), Mat(), var_types2, Mat(), params, true );

}