edu.wlu.cs.levy.SNARLI
Class SOM

java.lang.Object
  |
  +--edu.wlu.cs.levy.SNARLI.Map2D
        |
        +--edu.wlu.cs.levy.SNARLI.SOM

public class SOM

extends Map2D

SOM implements Kohonen's Self-Organizing Map.

Since:

JDK1.2

Constructor Summary

SOM(double[][] X, int N1, int N2)
Builds SOM with random initial weights and uniform probability density function.

SOM(double[][] X, int N1, int N2, double[] P)
Builds SOM with random initial weights.

SOM(double[][] X, int N1, int N2, double[] P, long seed)
Builds SOM with random initial weights.

SOM(double[][] X, int N1, int N2, double[] P, java.util.Random rand)
Builds SOM with random initial weights.

SOM(double[][] X, int N1, int N2, long seed)
Builds SOM with random initial weights and uniform probability density function.

SOM(double[][] X, int N1, int N2, java.util.Random rand)
Builds SOM with random initial weights and uniform probability density function.

Method Summary

void	learn(double ei, double ef, double sigma_i, double sigma_f, int tmax) Runs learning iterations.

Methods inherited from class edu.wlu.cs.levy.SNARLI.Map2D

getColumnDimension, getRowDimension, getUnit, getUnits, getWeights

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

SOM

public SOM(double[][] X,
           int N1,
           int N2,
           double[] P,
           java.util.Random rand)
    throws java.lang.IllegalArgumentException

Builds SOM with random initial weights.

Parameters:

X - data to learn

N1 - number of rows

N2 - number of columns

P - probability density function over X

rand - random-number generator

Throws:

java.lang.IllegalArgumentException - if X has fewer than two dimensions

java.lang.IllegalArgumentException - if X and P have different lengths

java.lang.IllegalArgumentException - if elements of P do not sum to 1

SOM

public SOM(double[][] X,
           int N1,
           int N2,
           double[] P)
    throws java.lang.IllegalArgumentException

Builds SOM with random initial weights.

Parameters:

X - data to learn

P - probability density function over X

N1 - number of rows

N2 - number of columns

Throws:

java.lang.IllegalArgumentException - if x and P have different lengths

java.lang.IllegalArgumentException - if elements of P do not sum to 1

SOM

public SOM(double[][] X,
           int N1,
           int N2,
           double[] P,
           long seed)
    throws java.lang.IllegalArgumentException

Builds SOM with random initial weights.

Parameters:

X - data to learn

P - probability density function over X

N1 - number of rows

N2 - number of columns

seed - seed for random-number generator

Throws:

java.lang.IllegalArgumentException - if X and P have different lengths

java.lang.IllegalArgumentException - if elements of P do not sum to 1

SOM

public SOM(double[][] X,
           int N1,
           int N2)

Builds SOM with random initial weights and uniform probability density function.

Parameters:

X - data to learn

N1 - number of rows

N2 - number of columns

SOM

public SOM(double[][] X,
           int N1,
           int N2,
           long seed)

Builds SOM with random initial weights and uniform probability density function.

Parameters:

X - data to learn

N1 - number of rows

N2 - number of columns

seed - seed for random-number generator

SOM

public SOM(double[][] X,
           int N1,
           int N2,
           java.util.Random rand)

Builds SOM with random initial weights and uniform probability density function.

Parameters:

X - data to learn

N1 - number of rows

N2 - number of columns

rand - random-number generator

Method Detail

learn

public void learn(double ei,
                  double ef,
                  double sigma_i,
                  double sigma_f,
                  int tmax)

Runs learning iterations. Learning rate e(t) = ei(ef/ei)^{t/tmax}.

Parameters:

sigma_i - initial width parameter

sigma_f - final width parameter

tmax - total number of time-steps to run