Som.h

/*
 
MIT License
 
Copyright (c) 2017 FMI Open Development / Markus Peura, first.last@fmi.fi
 
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
 
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
 
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
 
*/
/*
Part of Rack development has been done in the BALTRAD projects part-financed
by the European Union (European Regional Development Fund and European
Neighbourhood Partnership Instrument, Baltic Sea Region Programme 2007-2013)
*/
#ifndef SOM_H_
#define SOM_H_
 
 
 
 
/*
 g++ drain/examples/Som-example.cpp drain/util/{Caster,Variable,Debug}.cpp drain/image/{Geometry,Coordinates,Image,FilePng}.cpp  -lpng   -o Som-example
 */
#include <iostream>
#include <vector>
#include <list>
#include <limits>
#include <sstream>
 
#include <stdlib.h>
 
#include "drain/Variable.h"
 
 
 
namespace drain {
 
// Consider 
typedef std::vector<double> SomVector;  // for operator(std::ostream &)
 
template <class T = std::vector<double> >
class Som {
 
public:
 
    Som(int width=0, int height=0);
 
 
    inline
    void setGeometry(int width, int height) {
        this->width = width;
        this->height = height;
        map.resize(height);
        for (int j = 0; j < height; ++j) {
            map[j].resize(width);
        }
    }
 
    void fill(const T & sample){  // rename fill ?
 
        for (int i=0; i<width; i++){
            for (int j=0; j<height; j++){
                map[j][i] = sample;
            }
        }
    }
 
 
    void initialise(void (* init)(T & x)) {
 
        for (int i=0; i<width; i++){
            for (int j=0; j<height; j++){
                (*init)(map[j][i]);
            }
        }
 
    }
 
    inline
    void setNeighbourhoodRadius(double r){
        radius2 = r*r;
    }
 
    inline
    void setLearningCoefficient(double c){
        learningCoefficient = c;
    }
 
 
 
    void setDistanceFunction(double (* d)(const T & x1, const T & x2)) {
        distanceFunction = d;
    }
 
    void setMixingFunction(void (* mix)(const T & x1, const T & x2, double coeff, T & m)) {
        mixingFunction = mix;
    }
 
 
    inline
    int getHeight() const {
        return height;
    }
 
    inline
    int getWidth() const {
        return width;
    }
 
 
    void findBestMatchingUnit(const T & sample, int & iBest, int & jBest){
 
        double d;
        double dBest = std::numeric_limits<double>::max();
 
        for (int j=0; j<height; j++){
            for (int i=0; i<width; i++){
                d = (*distanceFunction)( sample, map[j][i]);
                if (d < dBest){
                    dBest = d;
                    iBest = i;
                    jBest = j;
                }
            }
        }
    }
 
 
 
    void train(const T & x){
        int iBest;
        int jBest;
        findBestMatchingUnit(x, iBest, jBest);
        train(x, iBest, jBest);
        /*
        double f;
        for (int j=0; j<height; ++j){
            for (int i=0; i<width; ++i){
                f = defaultNeighborhoodFunction(i-iBest, j-jBest);
                (*mixingFunction)(map[j][i], x, f, map[j][i]);
            }
        }
         */
    };
 
    inline
    void train(const T &x, int iBest, int jBest){
        double f;
        for (int j=0; j<height; ++j){
            for (int i=0; i<width; ++i){
                f = defaultNeighborhoodFunction(i-iBest, j-jBest);
                (*mixingFunction)(map[j][i], x, f, map[j][i]);
            }
        }
    };
 
 
 
    void toStream(std::ostream &ostr) const{
 
        std::string delimiter = "";
 
        ostr << title << "\n";
        for (int j=0; j<height; j++){
            for (int i=0; i<width; i++){
                ostr << j << ',' << i  << ':';
                ostr << '\t' << map[j][i] << '\n';
                //delimiter = "";
                //result += "\n";
            };
            ostr << '\n';
        };
 
    };
 
 
    //static
    double defaultNeighborhoodFunction(const int & i, const int & j)  const {
        return radius2 / (radius2 + i*i + j*j);
    }
 
    static
    double defaultDistanceMetric(const T & x1, const T & x2);
 
    static
    void defaultMixingFunction(const T & x1, const T & x2, double coeff, T & m);
 
 
 
 
    std::vector< std::vector<T> > map;
 
protected:
 
 
    double (* distanceFunction)(const  T & x1, const T & x2);
 
 
    void (* mixingFunction)(const T & x1, const T & x2, double alpha, T & result);
 
 
    std::string title;
 
    // signed, because comparisons will typically involve signed integers
    int width;
 
    // signed, because comparisons will typically involve signed integers
    int height;
 
    double radius2;
 
    double learningCoefficient;
 
};
 
 
 
/*
    For vectors and str objects having 
 */
template <class T>
double euclideanDistance2(const T & x1, const T & x2) {
 
    double d, result = 0.0;
 
    for (size_t i = 0; i < x1.size(); ++i) {// todo iterate
        d = x1[i]-x2[i]; // todo dist();
        result += d*d;
    }
 
    return result;
 
}
 
template <class T>
void vectorMix(const T & x1, const T & x2, double coeff, T & m){ // todo iterate
 
    for (size_t k = 0; k < x1.size(); ++k)
        m[k] = (1.0-coeff)*x1[k] + coeff*x2[k];
 
}
 
template <>
Som<>::Som(int width, int height){
    setGeometry(width, height);
    setDistanceFunction( euclideanDistance2 );    // consider Som<double>
    setMixingFunction( vectorMix ); // consider Som<double>
}
 
 
 
template <class T>
void uniformRandomVector256(std::vector<T> & x) {
 
    for (size_t k = 0; k < x.size(); ++k)
        x[k] = static_cast<T>(rand() & 0xff);
    //x[k] = static_cast<T2>(rand() & 0xffff)/static_cast<T2>(0xf00);
 
}
 
 
 
 
 
// Utilities
/*
template <class T>
static
double Som<std::vector<T> >::defaultNeighborhoodFunction(const int & i, const int & j) const {
    return radius2 / (radius2 + i*i + j*j);
}
 */
 
 
 
/*
template <class T>
double Som<T>::defaultDistanceMetric(const T & x1, const T & x2) {
 
    double d, result = 0.0;
 
    for (int i = 0; i < x1.size(); ++i) {
        d = x1[i]-x2[i];
        result += d*d;
    }
 
    return result;
 
}
 */
 
//class Som< std::vector<T> >;
/*
template <class T>
double Som< std::vector<T> >::defaultDistanceMetric(const T & x1, const T & x2) {
    return 123.456;
}
 */
 
/*
template <class T>
void Som<T>::defaultMixingFunction(const T & x1, const T & x2, double coeff, T & m){
 
    for (int k = 0; k < x1.size(); ++k)
        m[k] = (1.0-coeff)*x1[k] + coeff*x2[k];
 
}
 */
 
 
 
template <class T>
//std::ostream & operator<<(std::ostream & ostr, const NeuralVector<T> & n){
std::ostream & operator<<(std::ostream & ostr, const std::vector<T> & n){
 
    for (typename std::vector<T>::const_iterator it = n.begin(); it !=
            n.end(); ++it)
        ostr << *it << ' ';
    //ostr << '\n';
    return ostr;
};
 
template <class T>
//std::ostream & operator<<(std::ostream & ostr, const NeuralVector<T> & n){
std::ostream & operator<<(std::ostream & ostr, const Som<T> & som){
    som.toStream(ostr);
    return ostr;
};
 
 
} // drain
 
 
 
/* namespace drain */
 
#endif /* SOM_H_ */
 
// Drain