lapack_linear_least_squares.hpp

/*
 * Copyright (c) 2006-2014, Visualization and Multimedia Lab,
 *                          University of Zurich <http://vmml.ifi.uzh.ch>,
 *                          Eyescale Software GmbH,
 *                          Blue Brain Project, EPFL
 *
 * This file is part of VMMLib <https://github.com/VMML/vmmlib/>
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 * Redistributions of source code must retain the above copyright notice, this
 * list of conditions and the following disclaimer.  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.  Neither the name of the Visualization and Multimedia
 * Lab, University of Zurich nor the names of its contributors may be used to
 * endorse or promote products derived from this software without specific prior
 * written permission.
 * 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.
 */
#ifndef __VMML__VMMLIB_LAPACK_LINEAR_LEAST_SQUARES__HPP__
#define __VMML__VMMLIB_LAPACK_LINEAR_LEAST_SQUARES__HPP__

#include <vmmlib/matrix.hpp>
#include <vmmlib/vector.hpp>
#include <vmmlib/exception.hpp>

#include <vmmlib/lapack_types.hpp>
#include <vmmlib/lapack_includes.hpp>

#include <string>

namespace vmml
{

// XYYZZZ
// X    = data type: S - float, D - double
// YY   = matrix type, GE - general, TR - triangular
// ZZZ  = function name

namespace lapack
{

//
//
// SGELS/DGELS
//
//


// parameter struct
template< typename float_t >
struct llsq_params_xgels
{
    char            trans;  // 'N'->A, 'T'->Atransposed
    lapack_int      m;      // number of rows,      M >= 0
    lapack_int      n;      // number of columns,   N >= 0
    lapack_int      nrhs;   // number of columns of B/X
    float_t*        a;      // input A
    lapack_int      lda;    // leading dimension of A (number of rows)
    float_t*        b;      // input B, output X
    lapack_int      ldb;    // leading dimension of b
    float_t*        work;   // workspace
    lapack_int      lwork;  // workspace size
    lapack_int      info;   // 'return' value

    friend std::ostream& operator << ( std::ostream& os,
        const llsq_params_xgels< float_t >& p )
    {
        os
            << " m "        << p.m
            << " n "        << p.n
            << " nrhs "     << p.nrhs
            << " lda "      << p.lda
            << " ldb "      << p.ldb
            << " lwork "    << p.lwork
            << " info "     << p.info
            << std::endl;
        return os;
    }

};

// call wrappers

#if 0
void dgels_(const char *trans, const int *M, const int *N, const int *nrhs,
    double *A, const int *lda, double *b, const int *ldb, double *work,
    const int * lwork, int *info);
#endif

template< typename float_t >
inline void
llsq_call_xgels( llsq_params_xgels< float_t >& )
{
    VMMLIB_ERROR( "not implemented for this type.", VMMLIB_HERE );
}

template<>
inline void
llsq_call_xgels( llsq_params_xgels< float >& p )
{
    sgels_(
        &p.trans,
        &p.m,
        &p.n,
        &p.nrhs,
        p.a,
        &p.lda,
        p.b,
        &p.ldb,
        p.work,
        &p.lwork,
        &p.info
    );
}

template<>
inline void
llsq_call_xgels( llsq_params_xgels< double >& p )
{
    dgels_(
        &p.trans,
        &p.m,
        &p.n,
        &p.nrhs,
        p.a,
        &p.lda,
        p.b,
        &p.ldb,
        p.work,
        &p.lwork,
        &p.info
    );

}


template< size_t M, size_t N, typename float_t >
struct linear_least_squares_xgels
{
    bool compute(
        const matrix< M, N, float_t >& A,
        const vector< M, float_t >& B,
        vector< N, float_t >& x );

    linear_least_squares_xgels();
    ~linear_least_squares_xgels();

    const lapack::llsq_params_xgels< float_t >& get_params(){ return p; };

    matrix< M, N, float_t >& get_factorized_A() { return _A; }

protected:
    matrix< M, N, float_t > _A;
    vector< M, float_t >    _b;

    llsq_params_xgels< float_t > p;

};



template< size_t M, size_t N, typename float_t >
bool
linear_least_squares_xgels< M, N, float_t >::compute(
    const matrix< M, N, float_t >& A,
    const vector< M, float_t >& B,
    vector< N, float_t >& x )
{
    _A = A;
    _b = B;

    llsq_call_xgels( p );

    // success
    if ( p.info == 0 )
    {
        for( size_t index = 0; index < N; ++index )
        {
            x( index ) = _b( index );
        }

        return true;
    }
    if ( p.info < 0 )
    {
        VMMLIB_ERROR( "xGELS - invalid argument.", VMMLIB_HERE );
    }
    else
    {
        std::cout << "A\n" << A << std::endl;
        std::cout << "B\n" << B << std::endl;

        VMMLIB_ERROR( "least squares solution could not be computed.",
            VMMLIB_HERE );
    }
    return false;
}



template< size_t M, size_t N, typename float_t >
linear_least_squares_xgels< M, N, float_t >::
linear_least_squares_xgels()
{
    p.trans = 'N';
    p.m     = M;
    p.n     = N;
    p.nrhs  = 1;
    p.a     = _A.array;
    p.lda   = M;
    p.b     = _b.array;
    p.ldb   = M;
    p.work  = new float_t();
    p.lwork = -1;

    // workspace query
    llsq_call_xgels( p );

    p.lwork = static_cast< lapack_int > ( p.work[0] );
    delete p.work;

    p.work = new float_t[ p.lwork ];
}



template< size_t M, size_t N, typename float_t >
linear_least_squares_xgels< M, N, float_t >::
~linear_least_squares_xgels()
{
    delete[] p.work;
}



//
//
// SGESV/DGESV
//
//

template< typename float_t >
struct llsq_params_xgesv
{
    lapack_int      n; // order of matrix A = M * N
    lapack_int      nrhs; // number of columns of B
    float_t*        a;   // input A, output P*L*U
    lapack_int      lda; // leading dimension of A (for us: number of rows)
    lapack_int*     ipiv; // pivot indices, integer array of size N
    float_t*        b;  // input b, output X
    lapack_int      ldb; // leading dimension of b
    lapack_int      info;

    friend std::ostream& operator << ( std::ostream& os,
        const llsq_params_xgesv< float_t >& p )
    {
        os
            << "n "         << p.n
            << " nrhs "     << p.nrhs
            << " lda "      << p.lda
            << " ldb "      << p.ldvt
            << " info "     << p.info
            << std::endl;
        return os;
    }

};


#if 0
/* Subroutine */ int dgesv_(integer *n, integer *nrhs, doublereal *a, integer
    *lda, integer *ipiv, doublereal *b, integer *ldb, integer *info);
#endif


template< typename float_t >
inline void
llsq_call_xgesv( llsq_params_xgesv< float_t >& )
{
    VMMLIB_ERROR( "not implemented for this type.", VMMLIB_HERE );
}


template<>
inline void
llsq_call_xgesv( llsq_params_xgesv< float >& p )
{
    sgesv_(
        &p.n,
        &p.nrhs,
        p.a,
        &p.lda,
        p.ipiv,
        p.b,
        &p.ldb,
        &p.info
    );

}


template<>
inline void
llsq_call_xgesv( llsq_params_xgesv< double >& p )
{
    dgesv_(
        &p.n,
        &p.nrhs,
        p.a,
        &p.lda,
        p.ipiv,
        p.b,
        &p.ldb,
        &p.info
    );
}


template< size_t M, size_t N, typename float_t >
struct linear_least_squares_xgesv
{
    // computes x ( Ax = b ). x replaces b on output.
    void compute(
        matrix< N, N, float_t >& A,
        matrix< N, M, float_t >& b
        );

    linear_least_squares_xgesv();
    ~linear_least_squares_xgesv();

    const lapack::llsq_params_xgesv< float_t >& get_params() { return p; }

    lapack::llsq_params_xgesv< float_t > p;

}; // struct lapack_linear_least_squares


template< size_t M, size_t N, typename float_t >
void
linear_least_squares_xgesv< M, N, float_t >::
compute(
        matrix< N, N, float_t >& A,
        matrix< N, M, float_t >& b
        )
{
    p.a = A.array;
    p.b = b.array;

    lapack::llsq_call_xgesv( p );

    if ( p.info != 0 )
    {
        if ( p.info < 0 )
            VMMLIB_ERROR( "invalid value in input matrix", VMMLIB_HERE );
        else
            VMMLIB_ERROR( "factor U is exactly singular, solution could not be computed.", VMMLIB_HERE );
    }
}



template< size_t M, size_t N, typename float_t >
linear_least_squares_xgesv< M, N, float_t >::
linear_least_squares_xgesv()
{
    p.n     = N;
    p.nrhs  = M;
    p.lda   = N;
    p.ldb   = N;
    p.ipiv = new lapack_int[ N ];

}



template< size_t M, size_t N, typename float_t >
linear_least_squares_xgesv< M, N, float_t >::
~linear_least_squares_xgesv()
{
    delete[] p.ipiv;
}


} // namespace lapack

} // namespace vmml

#endif