KLLowerTriangularInference.cpp

Go to the documentation of this file.

 /*
 * Copyright (c) The Shogun Machine Learning Toolbox
 * Written (w) 2014 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.
 *
 * Code adapted from
 * http://hannes.nickisch.org/code/approxXX.tar.gz
 * and Gaussian Process Machine Learning Toolbox
 * http://www.gaussianprocess.org/gpml/code/matlab/doc/
 * and the reference paper is
 * Challis, Edward, and David Barber.
 * "Concave Gaussian variational approximations for inference in large-scale Bayesian linear models."
 * International conference on Artificial Intelligence and Statistics. 2011.
 *
 * This code specifically adapted from function in approxKL.m and infKL.m
 */

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

#include <shogun/mathematics/eigen3.h>
#include <shogun/mathematics/Math.h>
#include <shogun/mathematics/Statistics.h>

using namespace Eigen;

namespace shogun
{

CKLLowerTriangularInference::CKLLowerTriangularInference() : CKLInference()
{
    init();
}

CKLLowerTriangularInference::CKLLowerTriangularInference(CKernel* kern,
        CFeatures* feat, CMeanFunction* m, CLabels* lab, CLikelihoodModel* mod)
        : CKLInference(kern, feat, m, lab, mod)
{
    init();
}

void CKLLowerTriangularInference::init()
{
    SG_ADD(&m_InvK_Sigma, "invk_Sigma",
        "K^{-1}Sigma'",
        MS_NOT_AVAILABLE);
    SG_ADD(&m_mean_vec, "mean_vec",
        "The mean vector generated from mean function",
        MS_NOT_AVAILABLE);
    SG_ADD(&m_log_det_Kernel, "log_det_kernel",
        "The Log-determinant of Kernel",
        MS_NOT_AVAILABLE);

    SG_ADD(&m_Kernel_LsD, "L_sqrt_D",
        "The L*sqrt(D) matrix, where L and D are defined in LDLT factorization on Kernel*sq(m_scale)",
        MS_NOT_AVAILABLE);
    SG_ADD(&m_Kernel_P, "Permutation_P",
        "The permutation sequence of P, where P are defined in LDLT factorization on Kernel*sq(m_scale)",
        MS_NOT_AVAILABLE);
    m_log_det_Kernel=0;
}

CKLLowerTriangularInference::~CKLLowerTriangularInference()
{
}

SGVector<float64_t> CKLLowerTriangularInference::get_diagonal_vector()
{
    return SGVector<float64_t>();
}

void CKLLowerTriangularInference::update_deriv()
{
}

void CKLLowerTriangularInference::update_init()
{
    Eigen::LDLT<Eigen::MatrixXd> ldlt=update_init_helper();
    MatrixXd Kernel_D=ldlt.vectorD();
    MatrixXd Kernel_L=ldlt.matrixL();
    m_Kernel_LsD=SGMatrix<float64_t>(m_ktrtr.num_rows, m_ktrtr.num_cols);
    m_Kernel_LsD.zero();
    Map<MatrixXd> eigen_Kernel_LsD(m_Kernel_LsD.matrix, m_Kernel_LsD.num_rows, m_Kernel_LsD.num_cols);
    eigen_Kernel_LsD.triangularView<Lower>()=Kernel_L*Kernel_D.array().sqrt().matrix().asDiagonal();
    m_log_det_Kernel=2.0*eigen_Kernel_LsD.diagonal().array().abs().log().sum();

    m_Kernel_P=SGVector<index_t>(m_ktrtr.num_rows);
    for (index_t i=0; i<m_Kernel_P.vlen; i++)
        m_Kernel_P[i]=i;
    Map<VectorXi> eigen_Kernel_P(m_Kernel_P.vector, m_Kernel_P.vlen);
    eigen_Kernel_P=ldlt.transpositionsP()*eigen_Kernel_P;

    m_mean_vec=m_mean->get_mean_vector(m_features);
}

MatrixXd CKLLowerTriangularInference::solve_inverse(MatrixXd eigen_A)
{
    Map<VectorXi> eigen_Kernel_P(m_Kernel_P.vector, m_Kernel_P.vlen);
    Map<MatrixXd> eigen_Kernel_LsD(m_Kernel_LsD.matrix, m_Kernel_LsD.num_rows, m_Kernel_LsD.num_cols);

    //re-construct the Permutation Matrix
    PermutationMatrix<Dynamic> P(m_Kernel_P.vlen);
    P.setIdentity();
    SGVector<index_t> tmp=m_Kernel_P.clone();
    for (index_t i=0; i<tmp.vlen; i++)
    {
        while(tmp[i]>i)
        {
            P.applyTranspositionOnTheLeft(i,tmp[i]);
            index_t idx=tmp[i];
            tmp[i]=tmp[idx];
            tmp[idx]=idx;
        }
    }
    P=P.transpose();
    //(P'LDL'P)\eigen_A
    MatrixXd tmp1=P*eigen_A;
    MatrixXd tmp2=eigen_Kernel_LsD.triangularView<Lower>().solve(tmp1);
    MatrixXd tmp3=eigen_Kernel_LsD.triangularView<Lower>().transpose().solve(tmp2);
    return P.transpose()*tmp3;
}

float64_t CKLLowerTriangularInference::get_derivative_related_cov(SGMatrix<float64_t> dK)
{
    Map<MatrixXd> eigen_dK(dK.matrix, dK.num_rows, dK.num_cols);
    Map<VectorXd> eigen_alpha(m_alpha.vector, m_mu.vlen);
    Map<VectorXd> eigen_mu(m_mu.vector, m_mu.vlen);
    Map<MatrixXd> eigen_InvK_Sigma(m_InvK_Sigma.matrix, m_InvK_Sigma.num_rows, m_InvK_Sigma.num_cols);

    //dnlZ(j)=0.5*sum(sum(dK.*(K\((eye(n)- (invK_V+alpha*m'))')),2),1);
    MatrixXd tmp1=eigen_InvK_Sigma+eigen_alpha*(eigen_mu.transpose());
    MatrixXd tmp2=(MatrixXd::Identity(m_ktrtr.num_rows,m_ktrtr.num_cols)-tmp1).transpose();
    MatrixXd tmp3=solve_inverse(tmp2);
    return 0.5*(tmp3.array()*eigen_dK.array()).sum();
}

void CKLLowerTriangularInference::update_approx_cov()
{
}

void CKLLowerTriangularInference::update_chol()
{
    update_Sigma();
    update_InvK_Sigma();

    m_L=SGMatrix<float64_t>(m_ktrtr.num_rows, m_ktrtr.num_cols);
    Map<MatrixXd> eigen_L(m_L.matrix, m_L.num_rows, m_L.num_cols);
    Map<MatrixXd> eigen_InvK_Sigma(m_InvK_Sigma.matrix, m_InvK_Sigma.num_rows, m_InvK_Sigma.num_cols);
    MatrixXd tmp2=(eigen_InvK_Sigma-MatrixXd::Identity(m_ktrtr.num_rows,m_ktrtr.num_cols)).transpose();

    eigen_L=solve_inverse(tmp2);
}

} /* namespace shogun */