org.deeplearning4j.rl4j.learning

Class Learning<O extends org.deeplearning4j.rl4j.space.Encodable,A,AS extends org.deeplearning4j.rl4j.space.ActionSpace<A>,NN extends NeuralNet>

java.lang.Object

org.deeplearning4j.rl4j.learning.Learning<O,A,AS,NN>

All Implemented Interfaces:

ILearning<O,A,AS>, NeuralNetFetchable<NN>, StepCountable

Direct Known Subclasses:

AsyncLearning, SyncLearning

public abstract class Learning<O extends org.deeplearning4j.rl4j.space.Encodable,A,AS extends org.deeplearning4j.rl4j.space.ActionSpace<A>,NN extends NeuralNet>
extends Object
implements ILearning<O,A,AS>, NeuralNetFetchable<NN>

Author:

rubenfiszel ([email protected]) on 7/27/16. Useful factorisations and helper methods for class inheriting ILearning. Big majority of training method should inherit this

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	Learning.InitMdp<O>

Nested classes/interfaces inherited from interface org.deeplearning4j.rl4j.learning.ILearning

ILearning.LConfiguration

Constructor Summary

Constructors

Constructor and Description
Learning(ILearning.LConfiguration conf)

Method Summary

Modifier and Type	Method and Description
protected abstract DataManager	getDataManager()
static <O extends org.deeplearning4j.rl4j.space.Encodable,A,AS extends org.deeplearning4j.rl4j.space.ActionSpace<A>> org.nd4j.linalg.api.ndarray.INDArray	getInput(MDP<O,A,AS> mdp, O obs)
org.nd4j.linalg.api.ndarray.INDArray	getInput(O obs)
static Integer	getMaxAction(org.nd4j.linalg.api.ndarray.INDArray vector)
abstract NN	getNeuralNet()
int	incrementEpoch()
int	incrementStep()
Learning.InitMdp<O>	initMdp()
static <O extends org.deeplearning4j.rl4j.space.Encodable,A,AS extends org.deeplearning4j.rl4j.space.ActionSpace<A>> Learning.InitMdp<O>	initMdp(MDP<O,A,AS> mdp, IHistoryProcessor hp)
static int[]	makeShape(int size, int[] shape)
static int[]	makeShape(int batch, int[] shape, int length)
protected void	setHistoryProcessor(IHistoryProcessor.Configuration conf)

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.deeplearning4j.rl4j.learning.ILearning

getConfiguration, getMdp, getPolicy, getStepCounter, train

Constructor Detail

Learning

public Learning(ILearning.LConfiguration conf)

Method Detail

getMaxAction

public static Integer getMaxAction(org.nd4j.linalg.api.ndarray.INDArray vector)

getInput

public static <O extends org.deeplearning4j.rl4j.space.Encodable,A,AS extends org.deeplearning4j.rl4j.space.ActionSpace<A>> org.nd4j.linalg.api.ndarray.INDArray getInput(MDP<O,A,AS> mdp,
                                                                                                                                                                          O obs)

initMdp

public static <O extends org.deeplearning4j.rl4j.space.Encodable,A,AS extends org.deeplearning4j.rl4j.space.ActionSpace<A>> Learning.InitMdp<O> initMdp(MDP<O,A,AS> mdp,
                                                                                                                                                        IHistoryProcessor hp)

makeShape

public static int[] makeShape(int size,
                              int[] shape)

makeShape

public static int[] makeShape(int batch,
                              int[] shape,
                              int length)

getDataManager

protected abstract DataManager getDataManager()

getNeuralNet

public abstract NN getNeuralNet()

Specified by:

getNeuralNet in interface NeuralNetFetchable<NN extends NeuralNet>

incrementStep

public int incrementStep()

incrementEpoch

public int incrementEpoch()

setHistoryProcessor

protected void setHistoryProcessor(IHistoryProcessor.Configuration conf)

getInput

public org.nd4j.linalg.api.ndarray.INDArray getInput(O obs)

initMdp

public Learning.InitMdp<O> initMdp()