separation.cpp

模糊聚類分析源碼。包含教學文件

/*
  
  Context       : Fuzzy Clustering Algorithms

  Author        : Frank Hoeppner, see also AUTHORS file 

  Description   : implementation of class module Separation
                  
  History       :
    

  Comment       : 
    This file was generated automatically. DO NOT EDIT.

  Copyright     : Copyright (C) 1999-2000 Frank Hoeppner

    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
*/
/*
  The University of Applied Sciences Oldenburg/Ostfriesland/Wilhelmshaven
  hereby disclaims all copyright interests in the program package `fc' 
  (tool package for fuzzy cluster analysis) written by Frank Hoeppner.
  
  Prof. Haass, President of Vice, 2000-Mar-10
*/

#ifndef Separation_SOURCE
#define Separation_SOURCE

/* configuration include */
#ifdef HAVE_CONFIG_H
/*
//FILETREE_IFDEF HAVE_CONFIG_H
*/
#include "config.h"
/*
//FILETREE_ENDIF
*/
#endif



// necessary includes
#include "Separation.hpp"

// data


// implementation
template < class ANALYSIS >
Separation< ANALYSIS >::Separation
  (
  Algorithm<ANALYSIS>* ap_alg
  )
  : mp_succ_alg(ap_alg)
  
  
  {
  
  }
template < class ANALYSIS >
Separation< ANALYSIS >::~Separation
  (
  )
  {
  FUNCLOG("~Separation");
  
  delete mp_succ_alg;
  }
template < class ANALYSIS >
void
Separation< ANALYSIS >::operator()
  (
  ANALYSIS& a_analysis
  )
  {
  FUNCLOG("Separation");
  
  const int c(a_analysis.option().number_prototypes());
  const int n(a_analysis.option().number_features());
  const real_type m(a_analysis.option().fuzzifier());
  //const int dim(a_analysis.option().data_dimension());
  invariant(c>0,"non-empty prototype set",SOURCELOC);
  invariant(n>0,"non-empty data set",SOURCELOC);

  {
    trace("contractive properties");

    tuple_type summ;    // sum of u^{m-1}
    tuple_type sum;     // sum of u^m
    tuple_type sump;    // sum of u^{m+1}
    tuple_type prec;
    int p;

    summ.adjust(c);
    sum.adjust(c);
    sump.adjust(c);
    prec.adjust(c);

    matrix_set_scalar(summ,0);
    matrix_set_scalar(sum,0);
    matrix_set_scalar(sump,0);
    matrix_set_scalar(prec,0);

    typename ANALYSIS::link_iter i_link(a_analysis.links().begin());
    for (
        typename ANALYSIS::data_iter i_data(a_analysis.data().begin());
        i_data != a_analysis.data().end();
        ++i_data
        )
      {
      p=0;

      typename ANALYSIS::link_iter i_backup(i_link),i_link2;
      for (
          typename ANALYSIS::prot_iter i_prot(a_analysis.prototypes().begin());
          i_prot != a_analysis.prototypes().end();
          ++i_prot
          )
        {
        real_type u((*i_link).membership());

        i_link2 = i_backup;
        real_type uisum(1.0-u);
        for (
            typename ANALYSIS::prot_iter i_prot2(a_analysis.prototypes().begin());
            i_prot2 != a_analysis.prototypes().end();
            ++i_prot2
            )
          {
          if (i_prot!=i_prot2)
            uisum += (*i_link2).membership()
                   * sqrt( (*i_link).squared_distance()/(*i_link2).squared_distance() );
          ++i_link2;
          }

        real_type up;
        up = pow(u,m-1.0); summ[p] += up;
        up *= u; sum[p] += up; prec[p] += up*uisum;
        up *= u; sump[p] += up;

        ++p; ++i_link;
        }
      }

    trace("u^(m-1)=",summ);
    trace("u^m    =",sum);
    trace("u^(m+1)=",sump);
    matrix_dec(summ,sump);
    matrix_div(summ,sum);
    matrix_scale(summ,2.0*m/(m-1.0));
    trace("a=",summ);
    trace("norm",sqrt(matrix_square_norm(summ)));

    matrix_scale(prec,2.0*m/(m-1.0));
    matrix_div(prec,sum);
    trace("precise=",prec);

    a_analysis.option().contractive()=sqrt(matrix_square_norm(prec));
  a_analysis.option().validity_measure()=a_analysis.option().contractive();
  }
  {
    trace("compactness/separation (CS)");

    if (c<2) 
      {
      a_analysis.option().separation() = POS_IMPOSSIBLE_RANGE; 
      }
    else
      {
      real_type sep(0.0);

      typename ANALYSIS::link_iter i_link(a_analysis.links().begin());
      for (
          typename ANALYSIS::data_iter i_data(a_analysis.data().begin());
          i_data != a_analysis.data().end();
          ++i_data
          )
        for (
            typename ANALYSIS::prot_iter i_prot(a_analysis.prototypes().begin());
            i_prot != a_analysis.prototypes().end();
            ++i_prot
            )
          {
          sep += SQR((*i_link).membership())*
                 matrix_square_distance((*i_data).datum(),(*i_prot).center());
          ++i_link;
          }
      trace("sum(u^2*d^2)",sep);
      invariant(i_link==a_analysis.links().end(),"link size matches",SOURCELOC);

      real_type minprotdist(POS_IMPOSSIBLE_RANGE);
      for (
          typename ANALYSIS::prot_iter i_prot(a_analysis.prototypes().begin());
          i_prot != a_analysis.prototypes().end();
          ++i_prot
          ) if ((*i_prot).type()==CLUSTER_PROTOTYPE)
        for (
            typename ANALYSIS::prot_iter i_prot2(a_analysis.prototypes().begin());
            i_prot2 != a_analysis.prototypes().end();
            ++i_prot2
            ) if ((*i_prot2).type()==CLUSTER_PROTOTYPE)
          {
          if (i_prot!=i_prot2)
            {
            minprotdist=min(minprotdist,
              matrix_square_distance((*i_prot).center(),(*i_prot2).center()));
            }
          }
      trace("minimum prototype dist",minprotdist);

      if (minprotdist>0) sep /= n * minprotdist; else sep=POS_IMPOSSIBLE_RANGE;
      trace("separation",sep);

      a_analysis.option().separation() = sep;
      a_analysis.option().validity_measure() = sep;
      }
  }
  (*mp_succ_alg)(a_analysis);
  
  }

// template instantiation



#endif // Separation_SOURCE