sphclus.h

Go to the documentation of this file.

/*

Copyright (c) 2003, Cornell University
All rights reserved.

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 Cornell University 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.

*/

#if !defined(_SPHCLUS_H)
#define    _SPHCLUS_H

#include    "general.h"
#include    "cluster.h"
#include    "vec.h"
#include    "exceptions.h"
#include     "extravec.h"

namespace CLUS
{

/**    Class that describes Spheric Clusters
 */    
class SphericCluster:    public Cluster
{

protected:
    /// the sum of Ai(Xj)^2
    double              s_Ai;

    /// the sum of Ai(Xj)^2*Xj
    Vector<double>        s_Ai_Xj;
    
    Vector<double>        L;

public:
    SphericCluster():Cluster(),s_Ai(0.0),s_Ai_Xj(),L()
    {}
    
    SphericCluster(int InDim, int OutDim)
            :Cluster(InDim, OutDim), s_Ai(0.0), s_Ai_Xj(InDim+OutDim),
            L(InDim+OutDim)
    {}
    
    SphericCluster(int InDim, int OutDim, int):
            Cluster(InDim, OutDim), s_Ai_Xj(0), L(InDim+OutDim)
    {}

    void CopyFrom(SphericCluster& aux)
    {
        weight=aux.weight;
        dimension=aux.dimension;
        L=aux.L;
        state=aux.state;
    }
    
    void RandomCluster(void)
    {
        int i;
        for(i=0;i<inDim;i++)
        {
            L[i]=RANDOM01FLOAT*2-1;
        }
        for(i=inDim;i<inDim+outDim;i++)
        {
            L[i]=-0.4+0.8*RANDOM01FLOAT;
        }
    }
    
    void InitializeOnPoint(const double* point)
    {
        // we add a small random value to avoid problems if two clusters
        // are initialy in the same point
        for(int i=0; i<inDim+outDim; i++)
            L[i]=point[i]+RANDOM01FLOAT*1.0e-5;
    }

    inline double InferDistance(const double* InputData)
    {
        dataCache=InputData;
        distanceInfer=0;
        for(int i=0;i<inDim;i++)
            distanceInfer+=pow2(dataCache[i]-L[i]);
        distanceInfer*=weight;
        if (distanceInfer==0.0)
            distanceInfer=10e-300;
        return distanceInfer;/* added with 10e-16 to avoid 0.0 distances */
    }
    
    inline double ClusDistance(const double* TrainData)
    {
        distanceClus=0;
        dataCache=TrainData;
        int i;
        for(i=0;i<inDim;i++)
            distanceClus+=pow2(dataCache[i]-L[i]);
        distanceInfer=distanceClus*weight;
        for(i=inDim;i<inDim+outDim;i++)
            distanceClus+=pow2(dataCache[i]-L[i]);
        distanceClus*=weight;
        if (distanceClus==0.0)
            distanceClus=10e-300;
        return distanceClus;
    }
    
    inline void HBeginIdentif(void)
    {
        dimension=0.0;
    }

    inline double CorrectAppartGrade(double Coef)
    {
        double Ai=1/(distanceClus*Coef);
        double Ai2=pow2(Ai);
        s_Ai+=Ai2;
        for(int i=0;i<inDim+outDim;i++)
            s_Ai_Xj[i]+=Ai2*dataCache[i];
        return Ai2*distanceClus;

    }
    
    inline double HCorrectAppartGrade(double Coef)
    {
        double ret=CorrectAppartGrade(Coef);
        dimension+=ret;
        return ret;
    }
    
    inline double AdjustPrototypes(void)
    {
        double oldX, distP=0.0;
        for(int i=0;i<inDim+outDim;i++)
        {
            oldX=L[i];
            L[i]=s_Ai_Xj[i]/s_Ai;
            s_Ai_Xj[i]=0.0;
            distP+=pow2(oldX-L[i]);
        }

        s_Ai=0.0;
        return sqrt(distP)/(inDim+outDim);
    }
    
    inline double HAdjustPrototypes(void)
    {
        dimension/=s_Ai;

        cerr << dimension << endl;

        return AdjustPrototypes();
    }

    inline void AdjustYwithYoverD(Vector<double>& Y)
    {
        for(int i=0; i<outDim; i++)
            Y[i]+=L[i+inDim]/distanceInfer;
    }
    
    inline void AdjustYwithYoverD(Vector<double>& Y, double dist)
    {
        for(int i=0; i<outDim; i++)
            Y[i]+=L[i+inDim]/dist;
    }
    
    double ComputeLastY(void)
    {
        return L[inDim+outDim-1];
    }
    
    static string TypeName(void)
    {
        return string("SphericCluster");
    }
    
    void SaveToStream(ostream& out)
    {
        // all we need to save is L, no marker between in and out is necesarry
        out << "[ ";
        for(int i=0; i<inDim+outDim; i++)
            out << L[i] << " ";
        out << " ] { " ;
        out << weight << " }" << endl;
    }
    
    void LoadFromStream(istream& in)
    {
        string s;
        in >> s;
        if (s != "[")
            throw ErrMsg("[ missing in sphericcluster");
        for(int i=0; i<inDim+outDim; i++)
            in >> L[i];
        in >> s;
        if (s != "]")
            throw ErrMsg("] missing in sphericcluster");
        in >> s;
        if (s!="{")
            throw ErrMsg("{ missing in sphericcluster");
        in >> weight;
        in >> s;
        if (s!="}")
            throw ErrMsg("} missing in sphericcluster");

    }
    
    bool Split(SphericCluster& right, SphericCluster& save, double minDim)
    {
        if (state==Final)
            return false;
        if (state==Dead)
            throw ErrMsg("Attempt to split a dead cluster");
        if (state==Splited)
            throw ErrMsg("Attempt to split a splited cluster");

        if (dimension <=minDim)
            return false;

        // save *this in save
        save.weight=weight;
        save.state=Splited;
        save.L=L;
        save.dimension=dimension;

        state=right.state=Splitable;
        weight=right.weight=weight/2; // splited clusters are two times smaller

        Vector<double> delta(L.dim());
        GenerateRandomVector2(delta);
        SetNormTo(delta,dimension);

        // this cluster is the left cluster
        // first the right cluster
        right.L=L+delta;
        Dif(L,L.dim(),delta);
        dimension=right.dimension=0.0;

        return true;
    }
    
    bool CheckSplit(SphericCluster& left, SphericCluster& right, double splitCrit)
    {

        cerr << dimension << " : " << left.dimension << " " << right.dimension << endl;

        if (dimension*splitCrit > left.dimension + right.dimension)
        {
            // we have a good split
            return true;
        }
        else
        {
            // we have a bad split
            right.state=Dead;
            // copy back original version
            left.weight=weight;
            left.state=Final; // we dont split this anymore
            left.L=L;
            left.dimension=0.0;

            return false;
        }
    }
};
}

#endif    /* _SPHCLUS_H */

---