GaussianARDSparseKernel.cpp

Go to the documentation of this file.

/*
 * Copyright (c) The Shogun Machine Learning Toolbox
 * Written (w) 2015 Wu Lin
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 * 1. Redistributions of source code must retain the above copyright notice, this
 *    list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright notice,
 *    this list of conditions and the following disclaimer in the documentation
 *    and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * The views and conclusions contained in the software and documentation are those
 * of the authors and should not be interpreted as representing official policies,
 * either expressed or implied, of the Shogun Development Team.
 *
 */

#include <shogun/machine/gp/GaussianARDSparseKernel.h>

#ifdef HAVE_LINALG_LIB
#include <shogun/mathematics/linalg/linalg.h>
#endif

using namespace shogun;

CGaussianARDSparseKernel::CGaussianARDSparseKernel() : CGaussianARDKernel()
{
    initialize_sparse_kernel();
}
void CGaussianARDSparseKernel::initialize_sparse_kernel()
{
}

CGaussianARDSparseKernel::~CGaussianARDSparseKernel()
{
}

#if defined(HAVE_EIGEN3) && defined(HAVE_LINALG_LIB)
using namespace Eigen;

CGaussianARDSparseKernel::CGaussianARDSparseKernel(int32_t size)
        : CGaussianARDKernel(size)
{
    initialize_sparse_kernel();
}

CGaussianARDSparseKernel::CGaussianARDSparseKernel(CDotFeatures* l,
        CDotFeatures* r, int32_t size)
        : CGaussianARDKernel(l, r, size)
{
    initialize_sparse_kernel();
}

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

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

SGVector<float64_t> CGaussianARDSparseKernel::get_parameter_gradient_diagonal(
        const TParameter* param, index_t index)
{
    REQUIRE(param, "Param not set\n");
    if (!strcmp(param->m_name, "inducing_features"))
        return CKernel::get_parameter_gradient_diagonal(param, index);
    else
        return CGaussianARDKernel::get_parameter_gradient_diagonal(param, index);
}

SGMatrix<float64_t> CGaussianARDSparseKernel::get_parameter_gradient(
        const TParameter* param, index_t index)
{
    REQUIRE(param, "Param not set\n");
    if (!strcmp(param->m_name, "inducing_features"))
    {
        REQUIRE(lhs, "Left features not set!\n");
        REQUIRE(rhs, "Right features not set!\n");
        REQUIRE(index>=0 && index<num_lhs,"Index (%d) is out of bound (%d)\n",
            index, num_rhs);
        int32_t idx_l=index;
        //Note that CDotKernel requires lhs and rhs are CDotFeatures pointers
        //This Kernel is a subclass of CDotKernel
        SGVector<float64_t> left_vec=get_feature_vector(idx_l, lhs);
        SGMatrix<float64_t> res(left_vec.vlen, num_rhs);

        lazy_update_weights();

        for (int32_t idx_r=0; idx_r<num_rhs; idx_r++)
        {
            SGVector<float64_t> right_vec=get_feature_vector(idx_r, rhs);
            Map<VectorXd> eigen_res_col_vec(res.get_column_vector(idx_r),left_vec.vlen);

            SGVector<float64_t> vec=linalg::add(left_vec, right_vec, 1.0, -1.0);
            float64_t scalar_weight=1.0;
            //column vector
            SGMatrix<float64_t> right=compute_right_product(vec, scalar_weight);
            Map<VectorXd> eigen_right_col_vec(right.matrix,right.num_rows);

            if (m_ARD_type==KT_SCALAR)
            {
                scalar_weight*=m_weights_raw[0];
                eigen_res_col_vec=eigen_right_col_vec*scalar_weight;
            }
            else
            {
                if(m_ARD_type==KT_DIAG)
                {
                    Map<VectorXd> eigen_weights(m_weights_raw.matrix, m_log_weights.vlen);
                    eigen_res_col_vec=eigen_right_col_vec.cwiseProduct(eigen_weights);
                }
                else if(m_ARD_type==KT_FULL)
                {
                    Map<MatrixXd> eigen_weights(m_weights_raw.matrix, m_weights_raw.num_rows, m_weights_raw.num_cols);
                    eigen_res_col_vec=eigen_weights*eigen_right_col_vec;
                }
                else
                {
                    SG_ERROR("Unsupported ARD type\n");
                }

            }
            for (index_t i=0; i<left_vec.vlen; i++)
                res(i,idx_r)*=-kernel(idx_l,idx_r);
        }
        return res;
    }
    else
    {
        return CGaussianARDKernel::get_parameter_gradient(param, index);
    }
}
#endif /* HAVE_LINALG_LIB */