KRR.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) 2006 Mikio L. Braun
 * Written (W) 1999-2009 Soeren Sonnenburg
 * Copyright (C) 1999-2009 Fraunhofer Institute FIRST and Max-Planck-Society
 */

#include <shogun/lib/config.h>

#ifdef HAVE_LAPACK
#include <shogun/regression/KRR.h>
#include <shogun/mathematics/lapack.h>
#include <shogun/mathematics/Math.h>

using namespace shogun;

CKRR::CKRR()
: CKernelMachine()
{
    alpha=NULL;
    tau=1e-6;
}

CKRR::CKRR(float64_t t, CKernel* k, CLabels* lab)
: CKernelMachine()
{
    tau=t;
    set_labels(lab);
    set_kernel(k);
    alpha=NULL;
}


CKRR::~CKRR()
{
    SG_FREE(alpha);
}

bool CKRR::train_machine(CFeatures* data)
{
    SG_FREE(alpha);

    ASSERT(labels);
    if (data)
    {
        if (labels->get_num_labels() != data->get_num_vectors())
            SG_ERROR("Number of training vectors does not match number of labels\n");
        kernel->init(data, data);
    }
    ASSERT(kernel && kernel->has_features());

    // Get kernel matrix
    SGMatrix<float64_t> kernel_matrix=kernel->get_kernel_matrix<float64_t>();
    int32_t n = kernel_matrix.num_cols;
    int32_t m = kernel_matrix.num_rows;
    ASSERT(kernel_matrix.matrix && m>0 && n>0);

    for(int32_t i=0; i < n; i++)
        kernel_matrix.matrix[i+i*n]+=tau;

    // Get labels
    if (!labels)
        SG_ERROR("No labels set\n");

    SGVector<float64_t> alpha_orig=labels->get_labels();

    alpha=CMath::clone_vector(alpha_orig.vector, alpha_orig.vlen);

    if (alpha_orig.vlen!=n)
    {
        SG_ERROR("Number of labels does not match number of kernel"
                " columns (num_labels=%d cols=%d\n", alpha_orig.vlen, n);
    }

    clapack_dposv(CblasRowMajor,CblasUpper, n, 1, kernel_matrix.matrix, n, alpha, n);

    SG_FREE(kernel_matrix.matrix);
    return true;
}

bool CKRR::load(FILE* srcfile)
{
    SG_SET_LOCALE_C;
    SG_RESET_LOCALE;
    return false;
}

bool CKRR::save(FILE* dstfile)
{
    SG_SET_LOCALE_C;
    SG_RESET_LOCALE;
    return false;
}

CLabels* CKRR::apply()
{
    ASSERT(kernel);

    // Get kernel matrix
    SGMatrix<float64_t> kernel_matrix=kernel->get_kernel_matrix<float64_t>();
    int32_t n = kernel_matrix.num_cols;
    int32_t m = kernel_matrix.num_rows;
    ASSERT(kernel_matrix.matrix && m>0 && n>0);

    SGVector<float64_t> Yh(n);

    // predict
    // K is symmetric, CblasColMajor is same as CblasRowMajor 
    // and used that way in the origin call:
    // dgemv('T', m, n, 1.0, K, m, alpha, 1, 0.0, Yh, 1);
    int m_int = (int) m;
    int n_int = (int) n;
    cblas_dgemv(CblasColMajor, CblasTrans, m_int, n_int, 1.0, (double*) kernel_matrix.matrix,
        m_int, (double*) alpha, 1, 0.0, (double*) Yh.vector, 1);

    SG_FREE(kernel_matrix.matrix);

    return new CLabels(Yh);
}

float64_t CKRR::apply(int32_t num)
{
    ASSERT(kernel);

    // Get kernel matrix
    // TODO: use get_kernel_column instead of computing the whole matrix!
    SGMatrix<float64_t> kernel_matrix=kernel->get_kernel_matrix<float64_t>();
    int32_t n = kernel_matrix.num_cols;
    int32_t m = kernel_matrix.num_rows;
    ASSERT(kernel_matrix.matrix && m>0 && n>0);

    float64_t Yh;

    // predict
    Yh = CMath::dot(kernel_matrix.matrix + m*num, alpha, m);

    SG_FREE(kernel_matrix.matrix);
    return Yh;
}

#endif