SOMTrainer (boone)

java.lang.Object
- boone.Trainer
- - boone.training.SOMTrainer

All Implemented Interfaces:

Storable, java.lang.Cloneable

Direct Known Subclasses:

LVQTrainer
```
public class SOMTrainer
extends Trainer
```
The unsupervised SOM training algorithm implementing a simple dot product SOM with a linearly decreasing learn rate and a linearly decreasing neighborhood radius. The default learn rate is 100.0 and is decreased according to eta = eta / (currentEpoch + 1). The initial neighborhood radius is decreased linearly so that after all epochs the radius is 1.0. The default initial neighborhood radius is 3.0.

Author:

ghowa, Helmut A. Mayer

Field Summary

Fields
Modifier and Type	Field and Description
`protected double`	`deltaRadius` The change of radius per epoch.
`protected double`	`radius` The current neighborhood radius.
`protected double`	`startRadius` The initial neighborhood radius.
`protected Topology`	`topology` The topology of the SOM.

Fields inherited from class boone.Trainer
currentEpoch, epochs, learnRate, net, shuffle, stepMode, testData, trainingData, trainingSignalGenerator

Constructor Summary

Constructors
Constructor and Description

SOMTrainer(Topology topology)
Creates the SOM Trainer.

Constructors
Constructor and Description
`SOMTrainer(Topology topology)` Creates the SOM Trainer.

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`protected double`	`calcAdaptation(BrainPart part)` Not used.
`SOMTrainer`	`clone()` A deep clone.
`protected void`	`endBatch()` Decreases the neighborhood radius.
`protected void`	`endTrain()` Labels the neurons with cluster names.
`void`	`fromXML(org.jdom2.Element element)` Reads attributes from XML.
`double`	`getStartRadius()`
`Topology`	`getTopology()` Returns the topology of the SOM to be trained.
`protected void`	`reset()` Resets the neighborhood radius to the initial one, and sets the linear radius change such that after all epochs the radius is 1.0.
`void`	`setNetwork(NeuralNet net)` Sets the network for trainer and topology.
`void`	`setStartRadius(double startRadius)`
`double`	`test()` Returns the error on the programs pattern set by comparing the cluster of the winning neuron to the cluster of the pattern.
`org.jdom2.Element`	`toXML(org.jdom2.Element element)` Writes specific attributes to XML tree.
`protected void`	`train(java.util.List<java.lang.Double> inputPattern, java.util.List<java.lang.Double> targetPattern)` Trains a single pattern once.
`protected void`	`train(Neuron neuron, java.util.List<java.lang.Double> inputPattern, double learnRate)` Updates the weights from all input neurons to the given map neuron and normalizes the new weights.

Methods inherited from class boone.Trainer
getBatchSize, getEpochs, getLearnRate, getNetwork, getRank, getTestData, getTrainingData, getTrainingSignalGenerator, getWinningNeuron, setBatchSize, setEpochs, setLearnRate, setShuffle, setStepMode, setTestData, setTrainingData, setTrainingSignalGenerator, test, toString, train, trainPattern, updateLinks, updateNeurons

Methods inherited from class java.lang.Object
equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

- Field Detail
  - topology
```
protected Topology topology
```
    The topology of the SOM.
  - startRadius
```
protected double startRadius
```
    The initial neighborhood radius.
  - radius
```
protected double radius
```
    The current neighborhood radius.
  - deltaRadius
```
protected double deltaRadius
```
    The change of radius per epoch.
- Constructor Detail
  - SOMTrainer
```
public SOMTrainer(Topology topology)
```
    Creates the SOM Trainer. At this time the associated net is not known. Could be improved, but maybe too complicated.
    
    Parameters:
    
    topology - the SOM topology
- Method Detail
  - clone
```
public SOMTrainer clone()
```
    A deep clone.
    
    Overrides:
    
    clone in class Trainer
  - reset
```
protected void reset()
```
    Resets the neighborhood radius to the initial one, and sets the linear radius change such that after all epochs the radius is 1.0.
    
    Specified by:
    
    reset in class Trainer
  - setNetwork
```
public void setNetwork(NeuralNet net)
```
    Sets the network for trainer and topology.
    
    Overrides:
    
    setNetwork in class Trainer
    
    Parameters:
    
    net - a neural net
  - getTopology
```
public Topology getTopology()
```
    Returns the topology of the SOM to be trained.
    
    Returns:
    
    the topology
  - getStartRadius
```
public double getStartRadius()
```
  - setStartRadius
```
public void setStartRadius(double startRadius)
```
  - train
```
protected void train(java.util.List<java.lang.Double> inputPattern,
                     java.util.List<java.lang.Double> targetPattern)
```
    Trains a single pattern once.
    
    Specified by:
    
    train in class Trainer
    
    Parameters:
    
    inputPattern - an input pattern
    
    targetPattern - a target pattern (may be null)
  - calcAdaptation
```
protected double calcAdaptation(BrainPart part)
```
    Not used.
    
    Specified by:
    
    calcAdaptation in class Trainer
    
    Parameters:
    
    part - the brain part to be adapted
    
    Returns:
    
    the adaptation value
  - endBatch
```
protected void endBatch()
```
    Decreases the neighborhood radius.
    
    Overrides:
    
    endBatch in class Trainer
  - endTrain
```
protected void endTrain()
```
    Labels the neurons with cluster names.
    
    Specified by:
    
    endTrain in class Trainer
  - test
```
public double test()
```
    Returns the error on the programs pattern set by comparing the cluster of the winning neuron to the cluster of the pattern.
    
    Overrides:
    
    test in class Trainer
    
    Returns:
    
    the classification error
  - toXML
```
public org.jdom2.Element toXML(org.jdom2.Element element)
```
    Writes specific attributes to XML tree.
    
    Specified by:
    
    toXML in interface Storable
    
    Overrides:
    
    toXML in class Trainer
    
    Parameters:
    
    element - the parent element
    
    Returns:
    
    the trainer element
  - fromXML
```
public void fromXML(org.jdom2.Element element)
```
    Reads attributes from XML.
    
    Specified by:
    
    fromXML in interface Storable
    
    Overrides:
    
    fromXML in class Trainer
    
    Parameters:
    
    element - a JDOM element
  - train
```
protected void train(Neuron neuron,
                     java.util.List<java.lang.Double> inputPattern,
                     double learnRate)
```
    Updates the weights from all input neurons to the given map neuron and normalizes the new weights. The specific weight update follows the "Dot-Product SOM" (Kohonen 2001: 115) m(t+1) = m(t)+ a'(t)x(t)/||m(t) + a'(t)x(t)|| where m is the map neuron vector, x the input vector, and a' the learn rate
    
    Parameters:
    
    neuron - a map neuron
    
    inputPattern - a pattern
    
    learnRate - the learn rate

Class SOMTrainer

Field Summary

Fields inherited from class boone.Trainer

Constructor Summary

Method Summary

Methods inherited from class boone.Trainer

Methods inherited from class java.lang.Object

Field Detail

topology

startRadius

radius

deltaRadius

Constructor Detail

SOMTrainer

Method Detail

clone

reset

setNetwork

getTopology

getStartRadius

setStartRadius

train

calcAdaptation

endBatch

endTrain

test

toXML

fromXML

train