nuroko.module

Class Normaliser

java.lang.Object

nuroko.module.AComponent

nuroko.module.Normaliser

All Implemented Interfaces:

Cloneable, Iterable<IComponent>, IComponent, IFunction<mikera.vectorz.AVector,mikera.vectorz.AVector>, IInput, IInputOutput, IInputState, IModule, IOutput, IOutputState, IParameterised, ISynthesiser, IThinker, ITrainable

public class Normaliser
extends AComponent

Field Summary

Fields

Modifier and Type	Field and Description
protected mikera.vectorz.Vector	input
protected mikera.vectorz.Vector	inputGradient
protected mikera.vectorz.Vector	output
protected mikera.vectorz.Vector	outputGradient

Method Summary

Modifier and Type	Method and Description
Normaliser	clone() Creates a clone of a module, including a deep copy of any mutable state.
static Normaliser	create(mikera.vectorz.AVector mean, mikera.vectorz.AVector stdev)
static Normaliser	create(int length, double mean, double stdev)
void	generate(mikera.vectorz.AVector input, mikera.vectorz.AVector output)
List<IComponent>	getComponents()
mikera.vectorz.AVector	getGradient() Return an AVector referencing the accumulated gradient in this model
mikera.vectorz.Vector	getInput()
mikera.vectorz.Vector	getInputGradient()
IInputState	getInputState()
mikera.vectorz.Vector	getOutput()
mikera.vectorz.Vector	getOutputGradient()
mikera.vectorz.AVector	getParameters() Return an AVector referring to the parameters in the model.
boolean	hasDifferentTrainingThinking()
void	setInput(mikera.vectorz.AVector input)
void	thinkInternal() Thinks within the scope of the component.
void	trainGradientInternal(double factor) Abstract method for training gradients.

Methods inherited from class nuroko.module.AComponent

applyConstraints, applyConstraintsInternal, generate, getComponent, getConstraints, getDefaultLossFunction, getDownStack, getInputLength, getLearnFactor, getModules, getOutputLength, getParameterLength, getUpStack, initRandom, isStochastic, isSynthesiser, iterator, setConstraint, setLearnFactor, setOutput, think, think, thinkInternalTraining, topComponent, train, train, trainSynth, trainSynth

Methods inherited from class java.lang.Object

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

Methods inherited from interface java.lang.Iterable

forEach, spliterator

Field Detail

input

protected final mikera.vectorz.Vector input

inputGradient

protected final mikera.vectorz.Vector inputGradient

output

protected final mikera.vectorz.Vector output

outputGradient

protected final mikera.vectorz.Vector outputGradient

Method Detail

create

public static Normaliser create(mikera.vectorz.AVector mean,
                                mikera.vectorz.AVector stdev)

create

public static Normaliser create(int length,
                                double mean,
                                double stdev)

thinkInternal

public void thinkInternal()

Description copied from interface: IComponent

Thinks within the scope of the component. Updates component's output.

getComponents

public List<IComponent> getComponents()

Specified by:

getComponents in interface IComponent

getParameters

public mikera.vectorz.AVector getParameters()

Description copied from interface: IParameterised

Return an AVector referring to the parameters in the model. Expected to be mutable.

Returns:

getGradient

public mikera.vectorz.AVector getGradient()

Description copied from interface: IParameterised

Return an AVector referencing the accumulated gradient in this model

Returns:

trainGradientInternal

public void trainGradientInternal(double factor)

Description copied from class: AComponent

Abstract method for training gradients. Assumes use of the current activation states for all inputs/outputs. Should overwrite inputGradient, and adjust gradient for all parameters. Should apply own internal learn rate factor

Specified by:

trainGradientInternal in interface IComponent

Specified by:

trainGradientInternal in class AComponent

clone

public Normaliser clone()

Description copied from interface: IModule

Creates a clone of a module, including a deep copy of any mutable state.

Specified by:

clone in interface IComponent

Specified by:

clone in interface IModule

Specified by:

clone in interface IParameterised

Specified by:

clone in interface ISynthesiser

Specified by:

clone in interface IThinker

Specified by:

clone in interface ITrainable

Overrides:

clone in class AComponent

Returns:

generate

public void generate(mikera.vectorz.AVector input,
                     mikera.vectorz.AVector output)

Specified by:

generate in interface ISynthesiser

Overrides:

generate in class AComponent

hasDifferentTrainingThinking

public boolean hasDifferentTrainingThinking()

getInputState

public IInputState getInputState()

Specified by:

getInputState in interface IComponent

Overrides:

getInputState in class AComponent

getInput

public mikera.vectorz.Vector getInput()

getOutput

public mikera.vectorz.Vector getOutput()

getOutputGradient

public mikera.vectorz.Vector getOutputGradient()

setInput

public void setInput(mikera.vectorz.AVector input)

Specified by:

setInput in interface IInputState

Overrides:

setInput in class AComponent

getInputGradient

public mikera.vectorz.Vector getInputGradient()