deepnetts.net.loss

Class CrossEntropyLoss

java.lang.Object

deepnetts.net.loss.CrossEntropyLoss

All Implemented Interfaces:

LossFunction, Serializable

public class CrossEntropyLoss
extends Object
implements LossFunction, Serializable

Average Cross Entropy Loss function commonly used for multi class classification problems. E = -1/n * SUM(SUM(t*ln(y))) Since its 1-of-n classification scheme, all outputs but target are zeros so it comes down to E = -1/n * SUM(ln(y_targetIdx))

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
CrossEntropyLoss(int layerWidth)
CrossEntropyLoss(NeuralNetwork neuralNet)

Method Summary

Modifier and Type	Method and Description
float[]	addPatternError(float[] actualOutput, float[] targetOutput) Calculates and returns error vector for specified actual and target outputs.
void	addRegularizationSum(float regSum) Adds specified regularization sum to total loss.
float	getTotal() Returns the total error calculated by this loss function.
void	reset() Resets the total error and pattern counter.

Methods inherited from class java.lang.Object

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

Methods inherited from interface deepnetts.net.loss.LossFunction

valueFor

Constructor Detail

CrossEntropyLoss

public CrossEntropyLoss(NeuralNetwork neuralNet)

CrossEntropyLoss

public CrossEntropyLoss(int layerWidth)

Method Detail

addPatternError

public float[] addPatternError(float[] actualOutput,
                               float[] targetOutput)

Calculates and returns error vector for specified actual and target outputs.

Specified by:

addPatternError in interface LossFunction

Parameters:

actualOutput - actual output from the neural network

targetOutput - target/desired output of the neural network

Returns:

error vector for specified actual and target outputs

addRegularizationSum

public void addRegularizationSum(float regSum)

Description copied from interface: LossFunction

Adds specified regularization sum to total loss.

Specified by:

addRegularizationSum in interface LossFunction

Parameters:

regSum - regularization sum

getTotal

public float getTotal()

Description copied from interface: LossFunction

Returns the total error calculated by this loss function.

Specified by:

getTotal in interface LossFunction

Returns:

total error calculated by this loss function

reset

public void reset()

Description copied from interface: LossFunction

Resets the total error and pattern counter.

Specified by:

reset in interface LossFunction