javax.datamining.algorithm.feedforwardneuralnet
Interface Backpropagation

All Superinterfaces:

LearningAlgorithm

public interface Backpropagation

extends LearningAlgorithm

A Backpropagation object specifies the parameters used by the backpropagation learning algorithm. Back progation is a popular gradient descent method for training neural networks.

In back propagation, the output error on the training examples is used to adjust the network weights. The weight update rule can be specified as: w(i + 1) = w(i) - (learningRate * errorOfWeight), where w is a weight, i is the iteration, learningRate is the learning rate, and errorOfWeight is the error for the weight as determined by the partial derivative of the error with respect to the weight. The learning rate must be chosen between 0 and 1.

Author:

JSR-73 Java Data Mining Expert Group

See Also:

LearningAlgorithm

Method Summary

double	getLearningRate() Get the learning rate associated with a Backpropagation method.
double	getMomentum() Returns the momentum to be used as an alternative gradient descent method.
void	setLearningRate(double rate) Set the learning rate for a Backpropagation method.
void	setMomentum(double momentum) Sets the momentum as an alternative gradient descent method.

Method Detail

getLearningRate

public double getLearningRate()

Get the learning rate associated with a Backpropagation method.

Returns:

double

getMomentum

public double getMomentum()

Returns the momentum to be used as an alternative gradient descent method. Returns null if not supported.

Returns:

double

setLearningRate

public void setLearningRate(double rate)

Set the learning rate for a Backpropagation method. The value of the rate must be between 0 and 1.

Parameters:

rate - The learning rate to be used by the learning algorithm.

Returns:

void

setMomentum

public void setMomentum(double momentum)

Sets the momentum as an alternative gradient descent method. The momentum value ranges between 0 and 1, noted as [0,1) where 1 is excluded.

Parameters:

momentum - The momentum of learning to be used by the learning algorithm.

Returns:

void