burlap.behavior.singleagent.learnfromdemo.apprenticeship

Class ApprenticeshipLearningRequest

java.lang.Object

burlap.behavior.singleagent.learnfromdemo.IRLRequest

burlap.behavior.singleagent.learnfromdemo.apprenticeship.ApprenticeshipLearningRequest

public class ApprenticeshipLearningRequest
extends IRLRequest

A data structure for setting all the parameters of Max Margin Apprenticeship learning.

Author:

Stephen Brawner and Mark Ho

Field Summary

Fields

Modifier and Type	Field and Description
static double	DEFAULT_EPSILON
static int	DEFAULT_MAXITERATIONS
static int	DEFAULT_POLICYCOUNT
static boolean	DEFAULT_USEMAXMARGIN
protected double	epsilon The maximum feature score to cause termination of Apprenticeship learning
protected StateToFeatureVectorGenerator	featureGenerator The state feature generator that turns a state into a feature vector on which the reward function is assumed to be modeled
protected int	maxIterations The maximum number of iterations of apprenticeship learning
protected int	policyCount The maximum number of times a policy is rolled out and evaluated
protected StateGenerator	startStateGenerator The initial state generator that models the initial states from which the expert trajectories were drawn
protected double[]	tHistory the history of scores across each reward function improvement
protected boolean	useMaxMargin If true, use the full max margin method (expensive); if false, use the cheaper projection method

Fields inherited from class burlap.behavior.singleagent.learnfromdemo.IRLRequest

domain, expertEpisodes, gamma, planner

Constructor Summary

Constructors

Constructor and Description
ApprenticeshipLearningRequest()
ApprenticeshipLearningRequest(Domain domain, Planner planner, StateToFeatureVectorGenerator featureGenerator, java.util.List<EpisodeAnalysis> expertEpisodes, StateGenerator startStateGenerator)

Method Summary

Methods

Modifier and Type	Method and Description
Domain	getDomain()
double	getEpsilon()
java.util.List<EpisodeAnalysis>	getExpertEpisodes()
StateToFeatureVectorGenerator	getFeatureGenerator()
double	getGamma()
int	getMaxIterations()
Planner	getPlanner()
int	getPolicyCount()
StateGenerator	getStartStateGenerator()
double[]	getTHistory()
boolean	getUsingMaxMargin()
protected void	initDefaults()
boolean	isValid() Returns true if this request object has valid data members set; false otherwise.
void	setDomain(Domain d)
void	setEpsilon(double epsilon)
void	setExpertEpisodes(java.util.List<EpisodeAnalysis> episodeList)
void	setFeatureGenerator(StateToFeatureVectorGenerator stateFeaturesGenerator)
void	setGamma(double gamma)
void	setMaxIterations(int maxIterations)
void	setPlanner(Planner p)
void	setPolicyCount(int policyCount)
void	setStartStateGenerator(StateGenerator startStateGenerator)
void	setTHistory(double[] tHistory)
void	setUsingMaxMargin(boolean useMaxMargin)

Methods inherited from class java.lang.Object

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

Field Detail

featureGenerator

protected StateToFeatureVectorGenerator featureGenerator

The state feature generator that turns a state into a feature vector on which the reward function is assumed to be modeled

startStateGenerator

protected StateGenerator startStateGenerator

The initial state generator that models the initial states from which the expert trajectories were drawn

epsilon

protected double epsilon

The maximum feature score to cause termination of Apprenticeship learning

maxIterations

protected int maxIterations

The maximum number of iterations of apprenticeship learning

policyCount

protected int policyCount

The maximum number of times a policy is rolled out and evaluated

tHistory

protected double[] tHistory

the history of scores across each reward function improvement

useMaxMargin

protected boolean useMaxMargin

If true, use the full max margin method (expensive); if false, use the cheaper projection method

DEFAULT_EPSILON

public static final double DEFAULT_EPSILON

See Also:

Constant Field Values

DEFAULT_MAXITERATIONS

public static final int DEFAULT_MAXITERATIONS

See Also:

Constant Field Values

DEFAULT_POLICYCOUNT

public static final int DEFAULT_POLICYCOUNT

See Also:

Constant Field Values

DEFAULT_USEMAXMARGIN

public static final boolean DEFAULT_USEMAXMARGIN

See Also:

Constant Field Values

Constructor Detail

ApprenticeshipLearningRequest

public ApprenticeshipLearningRequest()

ApprenticeshipLearningRequest

public ApprenticeshipLearningRequest(Domain domain,
                             Planner planner,
                             StateToFeatureVectorGenerator featureGenerator,
                             java.util.List<EpisodeAnalysis> expertEpisodes,
                             StateGenerator startStateGenerator)

Method Detail

initDefaults

protected void initDefaults()

isValid

public boolean isValid()

Description copied from class: IRLRequest

Returns true if this request object has valid data members set; false otherwise.

Overrides:

isValid in class IRLRequest

Returns:

true if this request object has valid data members set; false otherwise.

setDomain

public void setDomain(Domain d)

Overrides:

setDomain in class IRLRequest

setPlanner

public void setPlanner(Planner p)

Overrides:

setPlanner in class IRLRequest

setFeatureGenerator

public void setFeatureGenerator(StateToFeatureVectorGenerator stateFeaturesGenerator)

setExpertEpisodes

public void setExpertEpisodes(java.util.List<EpisodeAnalysis> episodeList)

Overrides:

setExpertEpisodes in class IRLRequest

setStartStateGenerator

public void setStartStateGenerator(StateGenerator startStateGenerator)

setGamma

public void setGamma(double gamma)

Overrides:

setGamma in class IRLRequest

setEpsilon

public void setEpsilon(double epsilon)

setMaxIterations

public void setMaxIterations(int maxIterations)

setPolicyCount

public void setPolicyCount(int policyCount)

setTHistory

public void setTHistory(double[] tHistory)

setUsingMaxMargin

public void setUsingMaxMargin(boolean useMaxMargin)

getDomain

public Domain getDomain()

Overrides:

getDomain in class IRLRequest

getPlanner

public Planner getPlanner()

Overrides:

getPlanner in class IRLRequest

getFeatureGenerator

public StateToFeatureVectorGenerator getFeatureGenerator()

getExpertEpisodes

public java.util.List<EpisodeAnalysis> getExpertEpisodes()

Overrides:

getExpertEpisodes in class IRLRequest

getStartStateGenerator

public StateGenerator getStartStateGenerator()

getGamma

public double getGamma()

Overrides:

getGamma in class IRLRequest

getEpsilon

public double getEpsilon()

getMaxIterations

public int getMaxIterations()

getPolicyCount

public int getPolicyCount()

getTHistory

public double[] getTHistory()

getUsingMaxMargin

public boolean getUsingMaxMargin()