burlap.behavior.singleagent.learnbydemo.mlirl.support

Class DifferentiableRF

java.lang.Object

burlap.behavior.singleagent.learnbydemo.mlirl.support.DifferentiableRF

All Implemented Interfaces:

RewardFunction

Direct Known Subclasses:

DiffVFRF, LinearDiffRFVInit, LinearStateActionDifferentiableRF, LinearStateDifferentiableRF

public abstract class DifferentiableRF
extends java.lang.Object
implements RewardFunction

An abstract class for defining differentiable reward functions.

Author:

James MacGlashan.

Field Summary

Fields

Modifier and Type	Field and Description
protected int	dim The parameter dimensionality
protected double[]	parameters The parameters of the reward functions.

Constructor Summary

Constructors

Constructor and Description
DifferentiableRF()

Method Summary

Methods

Modifier and Type	Method and Description
DifferentiableRF	copy() Creates a copy of this reward function.
protected abstract DifferentiableRF	copyHelper() A helper method for making a copy of this reward function.
abstract double[]	getGradient(State s, GroundedAction ga, State sp) Returns the gradient of the reward function for the given state transition.
int	getParameterDimension() Returns the parameter dimensionality
double[]	getParameters() Returns the parameters of this reward function.
void	randomizeParameters(double lowerVal, double upperVal, java.util.Random rand) Randomizes the parameter values using the given random number generator.
void	setParameter(int i, double p) Sets the value of a given parameter.
void	setParameters(double[] parameters)
java.lang.String	toString()

Methods inherited from class java.lang.Object

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

Methods inherited from interface burlap.oomdp.singleagent.RewardFunction

reward

Field Detail

parameters

protected double[] parameters

The parameters of the reward functions.

dim

protected int dim

The parameter dimensionality

Constructor Detail

DifferentiableRF

public DifferentiableRF()

Method Detail

getGradient

public abstract double[] getGradient(State s,
                   GroundedAction ga,
                   State sp)

Returns the gradient of the reward function for the given state transition.

Parameters:

s - the source state

ga - the action taken in the source state

sp - the resulting state from the action

Returns:

the gradient of the reward function for the given transition.

setParameters

public void setParameters(double[] parameters)

setParameter

public void setParameter(int i,
                double p)

Sets the value of a given parameter.

Parameters:

i - which parameter to set

p - the value of the parameter

getParameterDimension

public int getParameterDimension()

Returns the parameter dimensionality

Returns:

the parameter dimensionality

getParameters

public double[] getParameters()

Returns the parameters of this reward function. The parameters are *not* copied, so changes made externally to them affect this reward function.

Returns:

the parameters of this reward function

randomizeParameters

public void randomizeParameters(double lowerVal,
                       double upperVal,
                       java.util.Random rand)

Randomizes the parameter values using the given random number generator.

Parameters:

lowerVal - the lower parameter range value

upperVal - the upper parameter range value

rand - the random number generator to use

copyHelper

protected abstract DifferentiableRF copyHelper()

A helper method for making a copy of this reward function. THe parameters and dimensionality do not have to be copied, because they will be copied in the public copy() method.

Returns:

a copy of this reward function.

copy

public DifferentiableRF copy()

Creates a copy of this reward function. The parameters of the copy can be modified without affecting the parameters this source reward function.

Returns:

a copy of this reward function

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object