nuroko.module

Class AComponent

java.lang.Object

nuroko.module.AComponent

All Implemented Interfaces:

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

Direct Known Subclasses:

ACompoundComponent, ALayerStack, AOperationComponent, AWeightLayer, Bias, Normaliser, Sparsifier

public abstract class AComponent
extends Object
implements IComponent, Iterable<IComponent>

Constructor Summary

Constructors

Constructor and Description
AComponent()

Method Summary

Modifier and Type	Method and Description
void	applyConstraints()
protected void	applyConstraintsInternal()
AComponent	clone() Creates a clone of a module, including a deep copy of any mutable state.
mikera.vectorz.AVector	generate(mikera.vectorz.AVector output)
void	generate(mikera.vectorz.AVector input, mikera.vectorz.AVector output)
IComponent	getComponent(int i)
ArrayList<IConstraint>	getConstraints()
LossFunction	getDefaultLossFunction() Returns the default loss function that should be used
IComponent	getDownStack()
int	getInputLength()
IInputState	getInputState()
double	getLearnFactor()
List<IModule>	getModules() Returns a list of sub-components of this module
int	getOutputLength()
int	getParameterLength() Returns the length of the parameter vector for this model
IComponent	getUpStack()
void	initRandom() Randomly initialises a component's parameters
boolean	isStochastic()
boolean	isSynthesiser()
Iterator<IComponent>	iterator()
void	setConstraint(IConstraint con)
void	setInput(mikera.vectorz.AVector inputValues)
void	setLearnFactor(double value)
void	setOutput(mikera.vectorz.AVector outputValues)
mikera.vectorz.AVector	think(mikera.vectorz.AVector input)
void	think(mikera.vectorz.AVector input, mikera.vectorz.AVector output)
void	thinkInternalTraining() Thinks within the scope of the component.
IComponent	topComponent()
void	train(mikera.vectorz.AVector input, mikera.vectorz.AVector target) Trains towards a target value.
void	train(mikera.vectorz.AVector input, mikera.vectorz.AVector target, LossFunction loss, double factor) Trains towards a target value using a specified loss function.
abstract void	trainGradientInternal(double factor) Abstract method for training gradients.
void	trainSynth(mikera.vectorz.AVector input)
void	trainSynth(mikera.vectorz.AVector input, double factor)

Methods inherited from class java.lang.Object

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

Methods inherited from interface nuroko.core.IComponent

getComponents, hasDifferentTrainingThinking, thinkInternal

Methods inherited from interface nuroko.core.IParameterised

getGradient, getParameters

Methods inherited from interface nuroko.core.IInputState

getInput, getInputGradient

Methods inherited from interface nuroko.core.IOutputState

getOutput, getOutputGradient

Methods inherited from interface java.lang.Iterable

forEach, spliterator

Constructor Detail

AComponent

public AComponent()

Method Detail

topComponent

public IComponent topComponent()

think

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

Specified by:

think in interface IThinker

generate

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

Specified by:

generate in interface ISynthesiser

getUpStack

public IComponent getUpStack()

Specified by:

getUpStack in interface ISynthesiser

getDownStack

public IComponent getDownStack()

Specified by:

getDownStack in interface ISynthesiser

thinkInternalTraining

public void thinkInternalTraining()

Description copied from interface: IComponent

Thinks within the scope of the component. Update's component's output. May be different from regular thinking.

Specified by:

thinkInternalTraining in interface IComponent

Specified by:

thinkInternalTraining in interface ITrainable

think

public mikera.vectorz.AVector think(mikera.vectorz.AVector input)

Specified by:

think in interface IFunction<mikera.vectorz.AVector,mikera.vectorz.AVector>

Specified by:

think in interface IThinker

setConstraint

public void setConstraint(IConstraint con)

getConstraints

public ArrayList<IConstraint> getConstraints()

generate

public final mikera.vectorz.AVector generate(mikera.vectorz.AVector output)

Specified by:

generate in interface ISynthesiser

getDefaultLossFunction

public LossFunction getDefaultLossFunction()

Description copied from interface: IComponent

Returns the default loss function that should be used

Specified by:

getDefaultLossFunction in interface IComponent

Returns:

getLearnFactor

public final double getLearnFactor()

Specified by:

getLearnFactor in interface IComponent

setLearnFactor

public final void setLearnFactor(double value)

Specified by:

setLearnFactor in interface IComponent

train

public final void train(mikera.vectorz.AVector input,
                        mikera.vectorz.AVector target)

Description copied from interface: ITrainable

Trains towards a target value. Performs thinking as required. Updates parameter gradient Overwrites outputGradient and inputGradient in components

Specified by:

train in interface ITrainable

trainSynth

public final void trainSynth(mikera.vectorz.AVector input)

Specified by:

trainSynth in interface IComponent

Specified by:

trainSynth in interface ISynthesiser

trainSynth

public void trainSynth(mikera.vectorz.AVector input,
                       double factor)

Specified by:

trainSynth in interface ISynthesiser

train

public final void train(mikera.vectorz.AVector input,
                        mikera.vectorz.AVector target,
                        LossFunction loss,
                        double factor)

Description copied from interface: ITrainable

Trains towards a target value using a specified loss function. Updates parameter gradient Overwrites outputGradient and inputGradient in components

Specified by:

train in interface ITrainable

trainGradientInternal

public abstract void trainGradientInternal(double factor)

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

initRandom

public void initRandom()

Description copied from interface: IComponent

Randomly initialises a component's parameters

Specified by:

initRandom in interface IComponent

setInput

public void setInput(mikera.vectorz.AVector inputValues)

Specified by:

setInput in interface IInputState

setOutput

public void setOutput(mikera.vectorz.AVector outputValues)

Specified by:

setOutput in interface IOutputState

getParameterLength

public int getParameterLength()

Description copied from interface: IParameterised

Returns the length of the parameter vector for this model

Specified by:

getParameterLength in interface IParameterised

Returns:

getComponent

public IComponent getComponent(int i)

getModules

public List<IModule> getModules()

Description copied from interface: IModule

Returns a list of sub-components of this module

Specified by:

getModules in interface IModule

Returns:

getInputState

public IInputState getInputState()

Specified by:

getInputState in interface IComponent

getInputLength

public int getInputLength()

Specified by:

getInputLength in interface IInput

getOutputLength

public int getOutputLength()

Specified by:

getOutputLength in interface IOutput

isStochastic

public boolean isStochastic()

Specified by:

isStochastic in interface IComponent

isSynthesiser

public boolean isSynthesiser()

Specified by:

isSynthesiser in interface IComponent

applyConstraintsInternal

protected void applyConstraintsInternal()

applyConstraints

public final void applyConstraints()

Specified by:

applyConstraints in interface IComponent

clone

public AComponent 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 Object

iterator

public Iterator<IComponent> iterator()

Specified by:

iterator in interface Iterable<IComponent>