unbbayes.simulation.montecarlo.sampling
Class AMonteCarloSampling

java.lang.Object
  unbbayes.simulation.montecarlo.sampling.AMonteCarloSampling

All Implemented Interfaces:

IMonteCarloSampling, ILongTaskProgressObservable

Direct Known Subclasses:

MapMonteCarloSampling, MatrixMonteCarloSampling

public abstract class AMonteCarloSampling
extends Object
implements IMonteCarloSampling

Field Summary

protected int	currentProgress
protected String	currentProgressStatus
protected int[]	factors
protected int	maxProgress
protected int	nTrials
protected ProbabilisticNetwork	pn
protected Map<Integer,Integer>	sampledStatesMap
protected byte[][]	sampledStatesMatrix
protected List<Node>	samplingNodeOrderQueue
protected int[]	timesSampled

Constructor Summary

AMonteCarloSampling()

Method Summary

protected void	addToSamplingOrderQueue(ArrayList<Node> children, boolean[] nodeAddedList) Take the children of a node that have already been added to the queue.
protected void	computeFactors() Calculate the factors necessary to transform the linear coordinate into a multidimensional one (which is the the state for each possible node - target and evidence).
protected void	createSamplingOrderQueue() Creates the queue of the nodes that are going to be analyzed.
protected double[][]	getCumulativeDistributionFunction(double[] pmf) Creates the cumulative distribution function (cdf) based on the node RV's pmf.
int	getCurrentProgress() The current number of this long task progress.
String	getCurrentProgressStatus() Returns a message with a description of the current status of the long task progress.
protected Integer	getIndexInQueue(Node node) Retrieves the node's index in the queue.
int	getLinearCoord(int[] multidimensionalCoord) Get the linear coordinate from the multidimensional one.
int	getMaxProgress() The maximum number allowed for this long task progress.
byte[]	getMultidimensionalCoord(int linearCoord) Get the multidimensional coordinate from the linear one.
protected List<Integer>	getParentsIndexesInQueue(ProbabilisticNode node) Return the indexes (sampling order) in the queue for the parents of a given node.
int	getPercentageDone() Returns the percentage of the progress done so far.
protected double[]	getProbabilityMassFunction(int[] sampledStates, List<Integer> parentsIndexes, ProbabilisticNode node) Creates the probability mass function based on the states sampled for the parents.
abstract byte[][]	getSampledStatesCompactMatrix() Returns the generated compact sample matrix.
abstract Map<Integer,Integer>	getSampledStatesMap() Returns the generated sample map, with key = linear coord (representing the sates sampled) and value = number of times this key was sampled.
abstract byte[][]	getSampledStatesMatrix() Returns the generated sample matrix.
List<Node>	getSamplingNodeOrderQueue() Return the order the nodes are in the sampled matrix.
protected byte	getState(double[] pmf) Uses the pmf to retrieve the cdf to choose a state from a random generated number (between 0 and 1).
abstract int[]	getStatesSetTimesSampled() The number of times the ith set of states was sampled.
protected void	initSamplingOrderQueue(boolean[] nodeAddedList) Initializes the queue with the nodes that are root.
void	notityObservers(LongTaskProgressChangedEvent event)
void	registerObserver(ILongTaskProgressObserver observer)
void	removeObserver(ILongTaskProgressObserver observer)
abstract void	start(ProbabilisticNetwork pn, int nTrials) Generates the MC sample with the given size for the given probabilistic network.
protected void	updateProgress(int progress)
protected void	updateProgress(int progress, String progressStatus)

Methods inherited from class java.lang.Object

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

Field Detail

maxProgress

protected int maxProgress

currentProgress

protected int currentProgress

currentProgressStatus

protected String currentProgressStatus

pn

protected ProbabilisticNetwork pn

nTrials

protected int nTrials

samplingNodeOrderQueue

protected List<Node> samplingNodeOrderQueue

sampledStatesMap

protected Map<Integer,Integer> sampledStatesMap

sampledStatesMatrix

protected byte[][] sampledStatesMatrix

timesSampled

protected int[] timesSampled

factors

protected int[] factors

Constructor Detail

AMonteCarloSampling

public AMonteCarloSampling()

Method Detail

registerObserver

public void registerObserver(ILongTaskProgressObserver observer)

Specified by:

registerObserver in interface ILongTaskProgressObservable

removeObserver

public void removeObserver(ILongTaskProgressObserver observer)

Specified by:

removeObserver in interface ILongTaskProgressObservable

notityObservers

public void notityObservers(LongTaskProgressChangedEvent event)

Specified by:

notityObservers in interface ILongTaskProgressObservable

getMaxProgress

public int getMaxProgress()

Description copied from interface: ILongTaskProgressObservable

The maximum number allowed for this long task progress. It represents 100%.

Specified by:

getMaxProgress in interface ILongTaskProgressObservable

Returns:

the maximum number allowed for this long task progress. It represents 100%.

getCurrentProgress

public int getCurrentProgress()

Description copied from interface: ILongTaskProgressObservable

The current number of this long task progress. It represents a percentage.

Specified by:

getCurrentProgress in interface ILongTaskProgressObservable

Returns:

the current number of this long task progress. It represents a percentage.

getPercentageDone

public int getPercentageDone()

Description copied from interface: ILongTaskProgressObservable

Returns the percentage of the progress done so far. It should be Math.round((float)currentProgress / maxProgress * 10000), a number between 0 and 10000.

Specified by:

getPercentageDone in interface ILongTaskProgressObservable

Returns:

the percentage of the progress done so far.

getCurrentProgressStatus

public String getCurrentProgressStatus()

Description copied from interface: ILongTaskProgressObservable

Returns a message with a description of the current status of the long task progress.

Specified by:

getCurrentProgressStatus in interface ILongTaskProgressObservable

Returns:

a message with a description of the current status of the long task progress.

updateProgress

protected void updateProgress(int progress,
                              String progressStatus)

updateProgress

protected void updateProgress(int progress)

getSampledStatesMatrix

public abstract byte[][] getSampledStatesMatrix()

Returns the generated sample matrix. The row represents the ith trial and the column represents the jth node from the sampled order. The value matrix[i][j] represents the sampled state index (respecting the node's states order) for the jth node in the ith trial.

Specified by:

getSampledStatesMatrix in interface IMonteCarloSampling

Returns:

The generated sample matrix.

getSampledStatesCompactMatrix

public abstract byte[][] getSampledStatesCompactMatrix()

Returns the generated compact sample matrix. The row represents the ith sampled state set and the column represents the jth node from the sampled order. The value matrix[i][j] represents the sampled state index (respecting the node's states order) for the jth node in the ith sampled state set. To get the number of times the ith set of states was sampled use getStatesSetTimesSampled().

Specified by:

getSampledStatesCompactMatrix in interface IMonteCarloSampling

Returns:

The generated compact sample matrix.

getStatesSetTimesSampled

public abstract int[] getStatesSetTimesSampled()

The number of times the ith set of states was sampled.

Specified by:

getStatesSetTimesSampled in interface IMonteCarloSampling

Returns:

The number of times the ith set of states was sampled.

getSampledStatesMap

public abstract Map<Integer,Integer> getSampledStatesMap()

Returns the generated sample map, with key = linear coord (representing the sates sampled) and value = number of times this key was sampled.

Specified by:

getSampledStatesMap in interface IMonteCarloSampling

Returns:

The generated sample map.

start

public abstract void start(ProbabilisticNetwork pn,
                           int nTrials)

Generates the MC sample with the given size for the given probabilistic network.

Specified by:

start in interface IMonteCarloSampling

Parameters:

pn - Probabilistic network that will be used for sampling.

nTrials - Number of trials to generate.

getSamplingNodeOrderQueue

public List<Node> getSamplingNodeOrderQueue()

Return the order the nodes are in the sampled matrix.

Specified by:

getSamplingNodeOrderQueue in interface IMonteCarloSampling

Returns:

The order the nodes are in the sampled matrix.

createSamplingOrderQueue

protected void createSamplingOrderQueue()

Creates the queue of the nodes that are going to be analyzed.

initSamplingOrderQueue

protected void initSamplingOrderQueue(boolean[] nodeAddedList)

Initializes the queue with the nodes that are root. In other words. It will put in the queue the nodes that do not have parents.

Parameters:

nodeAddedList - Keeps track of the nodes that have already been added to the queue (nodeAddedList[nodeIndex]=true).

addToSamplingOrderQueue

protected void addToSamplingOrderQueue(ArrayList<Node> children,
                                       boolean[] nodeAddedList)

Take the children of a node that have already been added to the queue. Analyze them one by one and add the child that is not in the queue yet.

Parameters:

children - Children of a node that is already in the queue.

nodeAddedList - Nodes that have already been added to the queue.

getParentsIndexesInQueue

protected List<Integer> getParentsIndexesInQueue(ProbabilisticNode node)

Return the indexes (sampling order) in the queue for the parents of a given node.

Parameters:

node - The node to retrieve the parents for finding the indexes.

Returns:

List of indexes (sampling order) of a node's parents in the queue.

getIndexInQueue

protected Integer getIndexInQueue(Node node)

Retrieves the node's index in the queue.

Parameters:

node -

Returns:

getState

protected byte getState(double[] pmf)

Uses the pmf to retrieve the cdf to choose a state from a random generated number (between 0 and 1).

Parameters:

pmf - The probability mass function for the node RV that we want to sample the state for.

Returns:

The sampled state for a given RV (based on its pmf).

getCumulativeDistributionFunction

protected double[][] getCumulativeDistributionFunction(double[] pmf)

Creates the cumulative distribution function (cdf) based on the node RV's pmf.

Parameters:

pmf - The probability mass function of the RV to calculate the cdf.

Returns:

The cumulative distribution function (cdf) for the given pmf.

getProbabilityMassFunction

protected double[] getProbabilityMassFunction(int[] sampledStates,
                                              List<Integer> parentsIndexes,
                                              ProbabilisticNode node)

Creates the probability mass function based on the states sampled for the parents.

Parameters:

sampledStates - The states (sampledStates[nodeIndex]) sampled for the nodes (nodeIndex).

parentsIndexes - The nodeIndex for each parent.

node - The node/RV to calculate the pmf.

Returns:

The probability mass function (pmf) of the node RV.

computeFactors

protected void computeFactors()

Calculate the factors necessary to transform the linear coordinate into a multidimensional one (which is the the state for each possible node - target and evidence). FactorForNode[i + 1] = ProductOf(NumberOfStates[i]), for all previous nodes (i).

getLinearCoord

public final int getLinearCoord(int[] multidimensionalCoord)

Get the linear coordinate from the multidimensional one. LinearCoord = SumOf(StateOf[i] * FactorOf[i]), for all possible nodes (i).

Specified by:

getLinearCoord in interface IMonteCarloSampling

Parameters:

multidimensionalCoord - Multidimensional coordinate (represented by the state for each node).

Returns:

The corresponding linear coordinate.

getMultidimensionalCoord

public final byte[] getMultidimensionalCoord(int linearCoord)

Get the multidimensional coordinate from the linear one.

Specified by:

getMultidimensionalCoord in interface IMonteCarloSampling

Parameters:

linearCoord - The linear coordinate.

Returns:

The corresponding multidimensional coordinate.