mkl test

gistfile1.txt

#include <shogun/base/init.h>
#include <shogun/classifier/mkl/MKLClassification.h>
#include <shogun/regression/svr/MKLRegression.h>
#include <shogun/regression/svr/LibSVR.h>
#include <shogun/regression/svr/SVRLight.h>
#include <shogun/features/Labels.h>
#include <shogun/lib/DataType.h>
#include <shogun/kernel/CustomKernel.h>
#include <shogun/kernel/CombinedKernel.h>
#include <shogun/features/SimpleFeatures.h>

using namespace shogun;

#define N 1000

int main(int argc, char** argv)
{
	init_shogun_with_defaults();

	CMKLRegression* regressor = new CMKLRegression();
	//CMKLClassification* regressor = new CMKLClassification();
	regressor->set_mkl_norm(1.0);
	regressor->set_solver_type(ST_GLPK);
	//CSVRLight* regressor = new CSVRLight();
	//CLibSVR* regressor = new CLibSVR();
	SG_REF(regressor);
	//regressor->parallel->set_num_threads(1);

	SGVector<float64_t> vec(N);
	for (int i=0; i<N; i++) vec.vector[i] = (i%2) ? (1.0) : (-1.0);
	CLabels* labels = new CLabels(vec);
	SG_REF(labels);

	SGMatrix<float64_t> matrix1(N,N);
	SGMatrix<float64_t> matrix2(N,N);

	for (int i=0; i<N; i++)
		for (int j=0; j<N; j++)
		{
			matrix1.matrix[i*N+j] = (i-j)*(i-j)/(1000.0*1000.0);
			matrix2.matrix[i*N+j] = (i-j)*(i-j)/(1000.0*1000.0);

			if (i==j)
			{
				matrix1.matrix[i*N+j]+=100;
				matrix2.matrix[i*N+j]+=100;
			}
		}

	CCustomKernel* kernel1 = new CCustomKernel(matrix1);
	CCustomKernel* kernel2 = new CCustomKernel(matrix2);

	matrix1.destroy_matrix();
	matrix2.destroy_matrix();

	CCombinedKernel* ckernel = new CCombinedKernel();
	SG_REF(ckernel);
	ckernel->append_kernel(kernel1);
	ckernel->append_kernel(kernel2);

	regressor->set_kernel(ckernel);
	regressor->set_labels(labels);
	regressor->train();
	regressor->apply();

	SG_UNREF(ckernel);
	SG_UNREF(regressor);
	SG_UNREF(labels);
	exit_shogun();
	return 0;
}