Apache Commons Math 3.6.1 · org.apache.commons.math3.ml.neuralnet.sofm.KohonenUpdateAction

Update formula for Kohonen's Self-Organizing Map.
The update method modifies the features w of the "winning" neuron and its neighbours according to the following rule:


 w_new = w_old + α e^{(-d / σ)} * (sample - w_old)

where

α is the current learning rate,
σ is the current neighbourhood size, and
d is the number of links to traverse in order to reach the neuron from the winning neuron.

This class is thread-safe as long as the arguments passed to the constructor are instances of thread-safe classes.
Each call to the update method will increment the internal counter used to compute the current values for

the learning rate, and
the neighbourhood size.

Consequently, the function instances that compute those values (passed to the constructor of this class) must take into account whether this class's instance will be shared by multiple threads, as this will impact the training process.

Field Summary

Modifier and Type	Field and Description
private final DistanceMeasure	distance Distance function.
private final LearningFactorFunction	learningFactor Learning factor update function.
private final NeighbourhoodSizeFunction	neighbourhoodSize Neighbourhood size update function.
private final AtomicLong	numberOfCalls Number of calls to `update(Network, double[])`.

Constructor Summary

Access	Constructor and Description
public	KohonenUpdateAction(DistanceMeasure Distance function. distance, LearningFactorFunction Learning factor update function. learningFactor, NeighbourhoodSizeFunction Neighbourhood size update function. neighbourhoodSize)

Method Summary

Modifier and Type	Method and Description
private boolean	Returns: `true` if the update succeeded, `true` if a concurrent update has been detected. attemptNeuronUpdate(Neuron Neuron to be updated. n, double[] Training data. features, double Learning factor. learningRate) Tries to update a neuron.
private double[]	Returns: the new values for the features. computeFeatures(double[] Current values of the features. current, double[] Training data. sample, double Learning factor. learningRate) Computes the new value of the features set.
private Neuron	Returns: the winning neuron. findAndUpdateBestNeuron(Network Network. net, double[] Sample data. features, double Current learning factor. learningRate) Searches for the neuron whose features are closest to the given sample, and atomically updates its features.
public long	Returns: the current number of calls. getNumberOfCalls() Retrieves the number of calls to the `update` method.
public void	update(Network Network. net, double[] Training data. features) Implements org.apache.commons.math3.ml.neuralnet.UpdateAction.update. Updates the network in response to the sample `features`.
private void	updateNeighbouringNeuron(Neuron Neuron to be updated. n, double[] Training data. features, double Learning factor. learningRate) Atomically updates the given neuron.

Inherited from java.lang.Object:: clone equals finalize getClass hashCode notify notifyAll toString wait wait wait

Field Detail

distance	back to summary
private final DistanceMeasure distance Distance function.

learningFactor	back to summary
private final LearningFactorFunction learningFactor Learning factor update function.

neighbourhoodSize	back to summary
private final NeighbourhoodSizeFunction neighbourhoodSize Neighbourhood size update function.

numberOfCalls	back to summary
private final AtomicLong numberOfCalls Number of calls to `update(Network, double[])`.

Constructor Detail

KohonenUpdateAction	back to summary
public KohonenUpdateAction(DistanceMeasure distance, LearningFactorFunction learningFactor, NeighbourhoodSizeFunction neighbourhoodSize) Parameters distance:DistanceMeasure Distance function. learningFactor:LearningFactorFunction Learning factor update function. neighbourhoodSize:NeighbourhoodSizeFunction Neighbourhood size update function.

Method Detail

attemptNeuronUpdate back to summary

attemptNeuronUpdate	back to summary
private boolean attemptNeuronUpdate(Neuron n, double[] features, double learningRate) Tries to update a neuron. Parameters n:Neuron Neuron to be updated. features:double[] Training data. learningRate:double Learning factor. Returns:boolean `true` if the update succeeded, `true` if a concurrent update has been detected.

private boolean attemptNeuronUpdate(Neuron n, double[] features, double learningRate)

Tries to update a neuron.

Parameters

n:Neuron: Neuron to be updated.
features:double[]: Training data.
learningRate:double: Learning factor.

Returns:boolean

true if the update succeeded, true if a concurrent update has been detected.

computeFeatures	back to summary
private double[] computeFeatures(double[] current, double[] sample, double learningRate) Computes the new value of the features set. Parameters current:double[] Current values of the features. sample:double[] Training data. learningRate:double Learning factor. Returns:double[] the new values for the features.

computeFeatures

back to summary

private double[] computeFeatures(double[] current, double[] sample, double learningRate)

Computes the new value of the features set.

Parameters

current:double[]: Current values of the features.
sample:double[]: Training data.
learningRate:double: Learning factor.

Returns:double[]

the new values for the features.

findAndUpdateBestNeuron	back to summary
private Neuron findAndUpdateBestNeuron(Network net, double[] features, double learningRate) Searches for the neuron whose features are closest to the given sample, and atomically updates its features. Parameters net:Network Network. features:double[] Sample data. learningRate:double Current learning factor. Returns:Neuron the winning neuron.

findAndUpdateBestNeuron

back to summary

private Neuron findAndUpdateBestNeuron(Network net, double[] features, double learningRate)

Searches for the neuron whose features are closest to the given sample, and atomically updates its features.

Parameters

net:Network: Network.
features:double[]: Sample data.
learningRate:double: Current learning factor.

Returns:Neuron

the winning neuron.

getNumberOfCalls back to summary

getNumberOfCalls	back to summary
public long getNumberOfCalls() Retrieves the number of calls to the `update` method. Returns:long the current number of calls.

public long getNumberOfCalls()

Retrieves the number of calls to the update method.

Returns:long: the current number of calls.

update back to summary

update	back to summary
public void update(Network net, double[] features) Implements org.apache.commons.math3.ml.neuralnet.UpdateAction.update. Doc from org.apache.commons.math3.ml.neuralnet.UpdateAction.update. Updates the network in response to the sample `features`. Parameters net:Network Network. features:double[] Training data.

public void update(Network net, double[] features)

Implements org.apache.commons.math3.ml.neuralnet.UpdateAction.update.

Doc from org.apache.commons.math3.ml.neuralnet.UpdateAction.update.

Updates the network in response to the sample features.

Parameters

net:Network: Network.
features:double[]: Training data.

updateNeighbouringNeuron	back to summary
private void updateNeighbouringNeuron(Neuron n, double[] features, double learningRate) Atomically updates the given neuron. Parameters n:Neuron Neuron to be updated. features:double[] Training data. learningRate:double Learning factor.

updateNeighbouringNeuron

back to summary

private void updateNeighbouringNeuron(Neuron n, double[] features, double learningRate)

Atomically updates the given neuron.

Parameters

n:Neuron: Neuron to be updated.
features:double[]: Training data.
learningRate:double: Learning factor.

public Class KohonenUpdateAction

Field Summary

Constructor Summary

Method Summary

Field Detail

Constructor Detail

Method Detail