Package libai.nn.unsupervised

Class Kohonen

java.lang.Object

libai.nn.NeuralNetwork

libai.nn.unsupervised.UnsupervisedLearning

libai.nn.unsupervised.Kohonen

All Implemented Interfaces:

java.io.Serializable

public class Kohonen
extends UnsupervisedLearning

Kohonen's Self-organizative Maps or SOM or Kohonen. This maps are one of the most important unsupervised neural networks of the history. The most important feature of the kohonen's maps is the possibility of transform any multidimensional space into a R^2 space, providing a highly precise clustering method. One of the most famous examples for the kohonen's maps is the transform the RGB color cube into a plane where the reds, greens, blues, etc are clustered in a very similar way of the any color picker utility.

See Also:

Serialized Form

Field Summary

Fields inherited from class libai.nn.NeuralNetwork

plotter, progress, random

Constructor Summary

Constructors

Constructor	Description
Kohonen(int[] nperlayer, double neighborhood)	Constructor.
Kohonen(int[] nperlayer, double neighborhood, int[] neighboursX, int[] neighboursY)	Constructor.
Kohonen(int[] nperlayer, double neighborhood, int[] neighboursX, int[] neighboursY, java.util.Random rand)	Constructor.
Kohonen(int[] nperlayer, double neighborhood, java.util.Random random)	Constructor.

Method Summary

Modifier and Type	Method	Description
void	expandMap(Column[] patterns, Column[] answers, int offset, int length)	Label the output for the patterns and expand the results through the neighbors until the map is completely fill.
int[][]	getMap()
Column	getPrototypeAt(int i, int j)
Column	simulate(Column pattern)	Calculates the output for the pattern.
void	simulate(Column pattern, Column result)	Calculates the output for the pattern and left the result in result.
void	train(Column[] patterns, double alpha, int epochs, int offset, int length)	Train the map.

Methods inherited from class libai.nn.NeuralNetwork

error, error, euclideanDistance2, euclideanDistance2, gaussian, getDefaultRandomGenerator, getPlotter, getProgressBar, initializeProgressBar, open, open, open, save, setPlotter, setProgressBar, train, train

Methods inherited from class java.lang.Object

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

Methods inherited from class libai.nn.unsupervised.UnsupervisedLearning

train, validatePreconditions

Constructor Detail

Kohonen

public Kohonen(int[] nperlayer,
               double neighborhood,
               int[] neighboursX,
               int[] neighboursY)

Constructor. Creates a kohonen's map with nperlayer[0] inputs, nperlayer[1] rows and nperlayer[2] columns. Additional set the initial size of the neighborhood and the way in the neighbors are connected.

Parameters:

nperlayer - Number of neurons (input, rows and columns)

neighborhood - Initial size of the neighborhood

neighboursX - neighbors along the X-axis

neighboursY - neighbors along the Y-axis

Kohonen

public Kohonen(int[] nperlayer,
               double neighborhood,
               int[] neighboursX,
               int[] neighboursY,
               java.util.Random rand)

Constructor. Creates a kohonen's map with nperlayer[0] inputs, nperlayer[1] rows and nperlayer[2] columns. Additional set the initial size of the neighborhood and the way in the neighbors are connected.

Parameters:

nperlayer - Number of neurons (input, rows and columns)

neighborhood - Initial size of the neighborhood

neighboursX - neighbors along the X-axis as deltas from origin eg. +1, 0, -1, 2

neighboursY - neighbors along the Y-axis as deltas from origin eg. +1, 0, -1, 2

rand - Random generator used for creating matrices

Kohonen

public Kohonen(int[] nperlayer,
               double neighborhood)

Constructor. Creates a kohonen's map using the standard neighborhood (up, down, left, right). Alias of Kohonen(nperlayer, _neighborhood, new int[]{0,0,1,-1}, new int[]{-1,1,0,0});

Parameters:

nperlayer - Number of neurons (input, rows and columns)

neighborhood - Initial size of the neighborhood

Kohonen

public Kohonen(int[] nperlayer,
               double neighborhood,
               java.util.Random random)

Constructor. Creates a kohonen's map using the standard neighborhood (up, down, left, right). Alias of Kohonen(nperlayer, _neighborhood, new int[]{0,0,1,-1}, new int[]{-1,1,0,0});

Parameters:

nperlayer - Number of neurons (input, rows and columns)

neighborhood - Initial size of the neighborhood

random - Random generator used for creating matrices

Method Detail

train

public void train(Column[] patterns,
                  double alpha,
                  int epochs,
                  int offset,
                  int length)

Train the map. The answers are omitted for the training process but are necessary for the labeling of the map.

Specified by:

train in class UnsupervisedLearning

Parameters:

patterns - The patterns to be learned.

alpha - The learning rate.

epochs - The maximum number of iterations

offset - The first pattern position

length - How many patterns will be used.

simulate

public Column simulate(Column pattern)

Description copied from class: NeuralNetwork

Calculates the output for the pattern.

Specified by:

simulate in class NeuralNetwork

Parameters:

pattern - Pattern to use as input.

Returns:

The output for the neural network.

simulate

public void simulate(Column pattern,
                     Column result)

Description copied from class: NeuralNetwork

Calculates the output for the pattern and left the result in result.

Specified by:

simulate in class NeuralNetwork

Parameters:

pattern - Pattern to use as input.

result - The output for the input.

getPrototypeAt

public Column getPrototypeAt(int i,
                             int j)

getMap

public int[][] getMap()

Returns:

The label map.

expandMap

public void expandMap(Column[] patterns,
                      Column[] answers,
                      int offset,
                      int length)

Label the output for the patterns and expand the results through the neighbors until the map is completely fill. NOTE: The expansion isn't an standard process but is very helpful to avoid unknown answers.

Parameters:

patterns - The patterns to label

answers - The expected answer for the patterns

offset - The initial pattern position

length - How many patterns to label.