nuroko.module.loss

Class LossFunction

java.lang.Object

nuroko.module.loss.LossFunction

Direct Known Subclasses:

AbsoluteErrorLoss, CrossEntropyLoss, SquaredErrorLoss

public abstract class LossFunction
extends Object

Constructor Summary

Constructors

Constructor and Description
LossFunction()

Method Summary

Methods

Modifier and Type	Method and Description
abstract double	calculateError(mikera.vectorz.AVector output, mikera.vectorz.AVector target)
abstract void	calculateErrorDerivative(mikera.vectorz.AVector output, mikera.vectorz.AVector target, mikera.vectorz.AVector gradientOut) Calculates a loss function for the given output and target Overwrites the gradientOut vector with the error derivative.
void	calculateErrorDerivative(mikera.vectorz.AVector output, mikera.vectorz.AVector target, IComponent comp) Calculates an error derivative, storing it in the output gradient of the given component

Methods inherited from class java.lang.Object

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

Constructor Detail

LossFunction

public LossFunction()

Method Detail

calculateErrorDerivative

public abstract void calculateErrorDerivative(mikera.vectorz.AVector output,
                            mikera.vectorz.AVector target,
                            mikera.vectorz.AVector gradientOut)

Calculates a loss function for the given output and target Overwrites the gradientOut vector with the error derivative.

Parameters:

output -

target -

gradientOut - output paremeter (to be overwritten)

calculateErrorDerivative

public final void calculateErrorDerivative(mikera.vectorz.AVector output,
                            mikera.vectorz.AVector target,
                            IComponent comp)

Calculates an error derivative, storing it in the output gradient of the given component

calculateError

public abstract double calculateError(mikera.vectorz.AVector output,
                    mikera.vectorz.AVector target)