nca.cpp

Go to the documentation of this file.

/***************************************************************************
 *
 * Authors:    Carlos Oscar            coss@cnb.csic.es (2013)
 *
 * Unidad de  Bioinformatica of Centro Nacional de Biotecnologia , CSIC
 *
 * 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 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
 * 02111-1307  USA
 *
 *  All comments concerning this program package may be sent to the
 *  e-mail address 'xmipp@cnb.csic.es'
 ***************************************************************************/

#include "nca.h"
#include <data/numerical_tools.h>

void NeighbourhoodCA::setLabels(const Matrix1D<unsigned char> &labels)
{
    this->labels=labels;
}

void NeighbourhoodCA::setSpecificParameters(double lambda, int K)
{
    this->lambda=lambda;
    this->K=K;
}

double NeighbourhoodCA::objectiveFunction()
{
    // Project X onto the subspace given by A, Y=X*A
    matrixOperation_AB(*X,A,Y);

    // Compute the distance between the nearest neighbours of each observation
    // in this new subspace
    D2Y.resizeNoCopy(MAT_YSIZE(Y),K);
    FOR_ALL_ELEMENTS_IN_MATRIX2D(D2Y)
    {
        double aux=0.;
        int actualj=MAT_ELEM(idx,i,j);
        for (int l=0; l<MAT_XSIZE(Y); ++l)
        {
            double diff=MAT_ELEM(Y,i,l)-MAT_ELEM(Y,actualj,l);
            aux+=diff*diff;
        }
        MAT_ELEM(D2Y,i,j)=exp(-aux);
    }
    D2Y.rowSum(D2YRowSum);

    // Compute F
    double F=0.;
    for (int i=0; i<MAT_YSIZE(Y); ++i)
    {
        double sumClassi=0.;
        int labeli=VEC_ELEM(labels,i);
        for (int j=0; j<K; ++j)
        {
            int actualj=MAT_ELEM(idx,i,j);
            if (VEC_ELEM(labels,actualj)==labeli)
                sumClassi+=MAT_ELEM(D2Y,i,j);
        }
        if (fabs(VEC_ELEM(D2YRowSum,i))>1e-16)
            F+=sumClassi/VEC_ELEM(D2YRowSum,i);
    }

    // Compute regularization
    double sumA2=0.0;
    FOR_ALL_ELEMENTS_IN_MATRIX2D(A)
        sumA2+=MAT_ELEM(A,i,j)*MAT_ELEM(A,i,j);
    sumA2/=MAT_XSIZE(A)*MAT_YSIZE(A);

    return -(F-lambda*sumA2);
}

double ncaObjectiveFuntion(double *p, void *prm)
{
    auto *nca=(NeighbourhoodCA *)prm;
    Matrix2D<double> &A=nca->A;
    memcpy(&MAT_ELEM(A,0,0),&(p[1]),MAT_XSIZE(A)*MAT_YSIZE(A)*sizeof(double));
    double c=nca->objectiveFunction();
    return c;
}

void NeighbourhoodCA::reduceDimensionality()
{
    Matrix2D<double> D2;
    subtractColumnMeans(*X);
    kNearestNeighbours(*X, K, idx, D2, distance, false);

    size_t d=MAT_XSIZE(*X);
    A.initGaussian(d,outputDim,0,0.01);

    // Optimize A
    Matrix1D<double> pA(MAT_XSIZE(A)*MAT_YSIZE(A)), steps(MAT_XSIZE(A)*MAT_YSIZE(A));
    steps.initConstant(1);
    memcpy(&VEC_ELEM(pA,0),&MAT_ELEM(A,0,0),VEC_XSIZE(pA)*sizeof(double));
    double goal;
    int iter;
    powellOptimizer(pA,1,VEC_XSIZE(pA),&ncaObjectiveFuntion,this,0.001,goal,iter,steps,false);
    memcpy(&MAT_ELEM(A,0,0),&VEC_ELEM(pA,0),VEC_XSIZE(pA)*sizeof(double));

    // Reduction
    matrixOperation_AB(*X,A,Y);
}