multiclassdistribution.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.

*/

// -*- C++ -*-

#if !defined _CLUS_MULTICLASSDISTRIBUTION_H
#define _CLUS_MULTICLASSDISTRIBUTION_H

#ifdef CLUS_USE_XML
#include "xml.h"
#endif

namespace CLUS
{

/** Base class for all distributions that can predict a discrete variable.
    The methods are virtual since not only one subclass is used in any given
    tree and is easier to let the virtual function mechanism kick in.
 */
class MulticlassDistribution
{
protected:
    /// the number of classes the predicted variable has
    int noClasses;

    /// the value of the test that is usually computed during StopLearning
    double statisticalTest; 
public:
    MulticlassDistribution(int NoClasses): noClasses(NoClasses)
    { }

    virtual ~MulticlassDistribution(void)
    { }

    /** Infer will use data to produce noClasses normalized probabilities
    into result. 
    */
    virtual void Infer(const double* cdata, const int* ddata, double* result)
    {
        for (int i=0; i<noClasses; i++)
            result[i]=1.0/noClasses;
    }

    /** MultiplicativeInfer uses data to produce probabilities and
    multiplies these probabilities with the ones in result.
    */
    virtual void MultiplicativeInfer(const double* cdata, const int* ddata, double* result)
    {
        // Implement just a uniform distribution regardless of the inputs
        for (int i=0; i<noClasses; i++)
            result[i]/=noClasses;
    }

    /** Initialize the sufficient statistics that are maintained.
     */
    virtual void StartLearning(void)
{ }

    /** Update the sufficient statistics according to the current input.
    Should be used if the class label is known for sure.

    @param cdata        contains values for the continuous variables
    @param ddata        for the discrete ones
    @param classLabel   known classification label
    @param weightSample used to give different importance
                        to the samples (magnifying glass effect).
    */
    virtual void LearnSample(const double* cdata, const int* ddata, int classLabel, double weightSample=1.0)
    { }

    /** Update the sufficient statistics according to the current input.
    Should be used if the class label cannot be determined with
    certainty.

    @param cdata        contains values for the continuous variables
    @param ddata        for the discrete ones
    @param classProbabilities     classification probabilities
    @param weightSample used to give different importance
                        to the samples (magnifying glass effect).
    */
    virtual void LearnSample(const double* cdata, const int* ddata, double classProbabilities, double weightSample=1.0)
    { }


    /** Uses the sufficient statistics to compute estimates of the parameters of the
    distribution.
    */
    virtual void StopLearning(void)
    { }
    
    /** Returns the log of p-value=1-cdf of the apropriate statistical test.
    In other words it returns the probability that randomly (no correlations between 
    input and classlabel) we do as well. The smaller the p-value the more predictive
    the distribution. The criterion depends on the distribution.
    This function should be called only after StopLearning
    */
    virtual double PValueStatisticalTest(void)
    {
        return log(1.0/noClasses);
    } // p-value for uniform distribution }

    void SaveToStream(ostream& output)
    {}

#ifdef CLUS_USE_XML
    /** Prints the distribution in a stream in XML */
    virtual void PrintToXmlStream(ostream& output)
    {
        output << "<MulticlassDistribution";
        PrintAttribute(output, "noClasses", noClasses);
        output << "/>" << endl;
    }
#endif

    /** @return true if the classLabel index has no significant apearance */
    virtual bool IsClassLabelAbsent(int index)
    {
        return false;
    }

};

}

#endif // _CLUS_MULTICLASSDISTRIBUTION_H

---