GaussianARDKernel.cpp

Go to the documentation of this file.

/*
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 3 of the License, or
 * (at your option) any later version.
 *
 * Written (W) 2012 Jacob Walker
 *
 * Adapted from WeightedDegreeRBFKernel.cpp
 */

#include <shogun/lib/common.h>
#include <shogun/kernel/GaussianARDKernel.h>
#include <shogun/features/Features.h>
#include <shogun/io/SGIO.h>

using namespace shogun;

CGaussianARDKernel::CGaussianARDKernel() : CLinearARDKernel()
{
    init();
}


CGaussianARDKernel::CGaussianARDKernel(int32_t size, float64_t width)
        : CLinearARDKernel(size), m_width(width)
{
    init();
}

CGaussianARDKernel::CGaussianARDKernel(CDenseFeatures<float64_t>* l,
        CDenseFeatures<float64_t>* r, int32_t size, float64_t width)
        : CLinearARDKernel(size), m_width(width)
{
    init();
}

bool CGaussianARDKernel::init(CFeatures* l, CFeatures* r)
{
    return CLinearARDKernel::init(l,r);
}

void CGaussianARDKernel::init()
{
    m_width=1.0;

    SG_ADD(&m_width, "width", "Kernel width", MS_AVAILABLE, GRADIENT_AVAILABLE);
}

CGaussianARDKernel::~CGaussianARDKernel()
{
}

CGaussianARDKernel* CGaussianARDKernel::obtain_from_generic(CKernel* kernel)
{
    if (kernel->get_kernel_type()!=K_GAUSSIANARD)
    {
        SG_SERROR("Provided kernel is not of type CGaussianARDKernel!\n");
    }

    /* since an additional reference is returned */
    SG_REF(kernel);
    return (CGaussianARDKernel*)kernel;
}

float64_t CGaussianARDKernel::compute(int32_t idx_a, int32_t idx_b)
{
    REQUIRE(lhs && rhs, "Features not set!\n")

    SGVector<float64_t> avec=
            ((CDenseFeatures<float64_t>*) lhs)->get_feature_vector(idx_a);
    SGVector<float64_t> bvec=
        ((CDenseFeatures<float64_t>*) rhs)->get_feature_vector(idx_b);

    REQUIRE(avec.vlen==bvec.vlen, "Number of right and left hand "
            "features must be the same\n");

    float64_t result=0;

    for (index_t i = 0; i < avec.vlen; i++)
        result += CMath::pow((avec[i]-bvec[i])*m_weights[i], 2);

    return CMath::exp(-result/m_width);
}

SGMatrix<float64_t> CGaussianARDKernel::get_parameter_gradient(
        const TParameter* param, index_t index)
{
    REQUIRE(lhs && rhs, "Features not set!\n")

    if (!strcmp(param->m_name, "weights"))
    {
        SGMatrix<float64_t> derivative=get_kernel_matrix();

        for (index_t j=0; j<num_lhs; j++)
        {
            for (index_t k=0; k<num_rhs; k++)
            {
                SGVector<float64_t> avec=
                    ((CDenseFeatures<float64_t>*) lhs)->get_feature_vector(j);
                SGVector<float64_t> bvec=
                    ((CDenseFeatures<float64_t>*) rhs)->get_feature_vector(k);

                REQUIRE(avec.vlen==bvec.vlen, "Number of right and left hand "
                        "features must be the same\n");

                float64_t element=compute(j,k);
                float64_t product=CMath::pow((avec[index]-bvec[index]), 2)
                        *(m_weights[index]/m_width);

                derivative(j,k)=-2*element*product;
            }
        }

        return derivative;
    }
    else if (!strcmp(param->m_name, "width"))
    {
        SGMatrix<float64_t> derivative(num_lhs, num_rhs);

        for (index_t j=0; j<num_lhs; j++)
        {
            for (index_t k=0; k<num_rhs; k++)
            {
                SGVector<float64_t> avec=
                    ((CDenseFeatures<float64_t>*) lhs)->get_feature_vector(j);
                SGVector<float64_t> bvec=
                    ((CDenseFeatures<float64_t>*) rhs)->get_feature_vector(k);

                REQUIRE(avec.vlen==bvec.vlen, "Number of right and left hand "
                        "features must be the same\n");

                float64_t result=0;

                for (index_t i=0; i<avec.vlen; i++)
                    result+=CMath::pow((avec[i]-bvec[i])*m_weights[i], 2);

                derivative(j,k)=CMath::exp(-result/m_width)*
                        result/(m_width*m_width);
            }
        }

        return derivative;
    }
    else
    {
        SG_ERROR("Can't compute derivative wrt %s parameter\n", param->m_name);
        return SGMatrix<float64_t>();
    }
}