Localiser.timeDelayLocalisers

Class MCMC

java.lang.Object

Localiser.timeDelayLocalisers.MCMC

All Implemented Interfaces:

TimeDelayLocaliserModel

Direct Known Subclasses:

MCMCHM

public class MCMC
extends java.lang.Object
implements TimeDelayLocaliserModel

Markov chain Monte Carlo (MCMC) time delay based localisation algorithm.

This is an advanaced and highly computationally intensive localisation algorithm based on MCMC methods. For a good description see; The Transit Light Curve (TLC) Project.I. Four Consecutive Transits of the Exoplanet XO-1b Matthew J. Holman1

Input variables are an ArrayList> of observed time delays, along with a corresponding error ArrayList> and ArrayList> of hydrophone positions. The time delay and delay errors arraylist will be exactly the same size.

Each row in the time delay ArrayList contains a group time delays calculated from N synchrnoised hydrophones in the corresponding ArrayList> row. This allows potential locaisation between large groups of unsynchronised hydrophones, a situation always encountered during target motions analysis.

Multiple MCMC chains can and should be run. These are executed on different threads to take advantage of multi core processing as much as possible. Even so a large number of chains or large number of observed delays can result in significant processing times.

Results are analysed for convergence and final locations packed into an MCMCTDResults class. Input Time Delays Time Delay Errors Hydrophone Array SampaleRate

Author:

Jamie Macaulay

Constructor Summary

Constructors

Constructor and Description
MCMC(java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> hydrophoneArray, java.util.ArrayList<java.util.ArrayList<java.lang.Double>> timeDelays, java.util.ArrayList<java.util.ArrayList<java.lang.Double>> timeDelayErrors, MCMCParams settings, float sampleRate, double speedOfSound)
MCMC(MCMCParams settings)

Method Summary

Methods

Modifier and Type	Method and Description
static java.util.ArrayList<java.util.ArrayList<java.lang.Double>>	addCrossCorError(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> tdErrors, double timeError, float sampleRate) Adds the cross correlation error to the time delay errors array.
MCMCTDResults	AnalyseMCMCResults(java.util.ArrayList<javax.vecmath.Point3f> chainJumps, java.util.ArrayList<java.lang.Double> Chi2, int analysisType) Take an ArrayList of the Markov chain jumps and chi-squared values of each jump and calculate source position.
MCMCTDResults	averageChains(java.util.ArrayList<MCMCTDResults> data)
java.lang.Boolean	changeSettings() Change any settings in the algorithm.
static double	chiSquared(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> observedTimeDelays, java.util.ArrayList<java.util.ArrayList<java.lang.Double>> chainPosTimeDelays, java.util.ArrayList<java.util.ArrayList<java.lang.Double>> errors) Calculates the chi^2 value between observed data and theoretical delays calculated for a certain point in three dimensional space.
void	clearResults() Clear the results from memory.
java.util.ArrayList<MCMCTDResults>	clusterAnalaysis(java.util.ArrayList<MCMCTDResults> meanArray) Perform a cluster analysis of results.
static java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>>	compressMCMCResults(java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> jumps, int compressSize) Reduces the number of MCMC jumps in order to reduce memory usage when displaying probability distributions in Java3D or other visualise components.
java.util.ArrayList<java.util.ArrayList<java.lang.Double>>	getChiJumps()
java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>>	getHydrophonePos()
java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>>	getJumps()
javax.vecmath.Point3f	getNewJumpPoint(javax.vecmath.Point3f ChainPos) Calculates a random new Co-Ordinate, 3D Cartesian or Cylindrical space;
java.util.ArrayList<MCMCTDResults>	getResults()
java.util.ArrayList<MCMCTDResults>	getResultsAllChains()
MCMCParams	getSettings()
static java.util.ArrayList<java.util.ArrayList<java.lang.Double>>	getTimeDelays(javax.vecmath.Point3f chainPos, java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> hydrophonePos, double speedOfSound) Calculate the theoretical time delays if a source was located at point3d ChainPos.
java.util.ArrayList<javax.vecmath.Point3d>	kMeansAnalysis(java.util.ArrayList<MCMCTDResults> meanArray) Sometimes multiple chains will converge to different final distributions.
java.lang.Object[]	mCMC(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> observedTimeDelays) Simple Markov Chain Monte Carlo simulation.
double	randomNGenerator()
void	runAlgorithm() Runs the localisation algorithm an creates a results.
void	setHydrophonePos(java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> hydrophonePos)
void	setSampleRate(float sampleRate)
void	setSoundSpeed(double speedOfSound)
void	setTimeDelays(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> timeDelays)
void	setTimeDelaysErrors(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> timeDelayErrors) Note the sample rate must be set before this function.
void	stop() Stop the localiser.

Methods inherited from class java.lang.Object

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

Constructor Detail

MCMC

public MCMC(java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> hydrophoneArray,
    java.util.ArrayList<java.util.ArrayList<java.lang.Double>> timeDelays,
    java.util.ArrayList<java.util.ArrayList<java.lang.Double>> timeDelayErrors,
    MCMCParams settings,
    float sampleRate,
    double speedOfSound)

MCMC

public MCMC(MCMCParams settings)

Method Detail

getNewJumpPoint

public javax.vecmath.Point3f getNewJumpPoint(javax.vecmath.Point3f ChainPos)

Calculates a random new Co-Ordinate, 3D Cartesian or Cylindrical space;

Parameters:

ChainPos -

Returns:

randomNGenerator

public double randomNGenerator()

chiSquared

public static double chiSquared(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> observedTimeDelays,
                java.util.ArrayList<java.util.ArrayList<java.lang.Double>> chainPosTimeDelays,
                java.util.ArrayList<java.util.ArrayList<java.lang.Double>> errors)

Calculates the chi^2 value between observed data and theoretical delays calculated for a certain point in three dimensional space.

Parameters:

observedTimeDelays- - the observed time delays.

chainPosTimeDelays- - the set of time delays to compare to the observed time delays

errors- - the measurement errors in the observed time delays.

Returns:

chi squared value for observed time delays compared to chain position time delays.

getTimeDelays

public static java.util.ArrayList<java.util.ArrayList<java.lang.Double>> getTimeDelays(javax.vecmath.Point3f chainPos,
                                                                       java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> hydrophonePos,
                                                                       double speedOfSound)

Calculate the theoretical time delays if a source was located at point3d ChainPos.

Parameters:

ChainPos- - theoretical position of the source.

hydrophonePos- - the positions of all the hydrophones within the array in Cartesian coOrdinates.

speedOfSound- - the speed of sound;

Returns:

The time delays that would result from a source at this location.

AnalyseMCMCResults

public MCMCTDResults AnalyseMCMCResults(java.util.ArrayList<javax.vecmath.Point3f> chainJumps,
                               java.util.ArrayList<java.lang.Double> Chi2,
                               int analysisType)

Take an ArrayList of the Markov chain jumps and chi-squared values of each jump and calculate source position. In order to make sure the burn in phase is not included, a certain initial percentage of jumps is ignored.

Parameters:

chainJumps -

Chi2 -

PercentagetoIgnore -

Returns:

mCMC

public java.lang.Object[] mCMC(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> observedTimeDelays)

Simple Markov Chain Monte Carlo simulation. A good description of how this process works can be found in: The Transit Light Curve (TLC) Project.I. Four Consecutive Transits of the Exoplanet XO-1b Matthew J. Holman1, Joshua N. Winn2, David W. Latham1, Francis T. O�Donovan3, David Charbonneau1,7....

Parameters:

observedTimeDelays -

Returns:

Object containing an arraylist of successful chain jumps and an arraylist containing the corresponding chi squared values of each successful jump

runAlgorithm

public void runAlgorithm()

Description copied from interface: TimeDelayLocaliserModel

Runs the localisation algorithm an creates a results. The results are dependednt on the localisation class.

Specified by:

runAlgorithm in interface TimeDelayLocaliserModel

clusterAnalaysis

public java.util.ArrayList<MCMCTDResults> clusterAnalaysis(java.util.ArrayList<MCMCTDResults> meanArray)

Perform a cluster analysis of results. The clustering algorithm used is determined by the MCMC localiser params.

kMeansAnalysis

public java.util.ArrayList<javax.vecmath.Point3d> kMeansAnalysis(java.util.ArrayList<MCMCTDResults> meanArray)

Sometimes multiple chains will converge to different final distributions. If this is the case the results must be clustered, otherwise the final answer will simply be the mean of the two different distributions. Here a 3D k-means clustering algorithm is used in x, y and z . First, nk random points are assigned in space based on the start dispersion. Next, all chains are clusters corresponding to their euclidian distance from these two random points. The random points then move to the centroid of their correpsonding clusters. The process repeats 'int iterations' times by which time it is likely convergence will have occurred. Clustering is attempted 'int attempts' times from different random positions.

Parameters:

meanArray- - the mean values of all chains;

Returns:

averageChains

public MCMCTDResults averageChains(java.util.ArrayList<MCMCTDResults> data)

addCrossCorError

public static java.util.ArrayList<java.util.ArrayList<java.lang.Double>> addCrossCorError(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> tdErrors,
                                                                          double timeError,
                                                                          float sampleRate)

Adds the cross correlation error to the time delay errors array.

Parameters:

tdErrors -

timeError -

sampleRate -

Returns:

setTimeDelays

public void setTimeDelays(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> timeDelays)

setTimeDelaysErrors

public void setTimeDelaysErrors(java.util.ArrayList<java.util.ArrayList<java.lang.Double>> timeDelayErrors)

Note the sample rate must be set before this function.

Parameters:

timeDelayErrors -

Throws:

java.lang.Exception

setHydrophonePos

public void setHydrophonePos(java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> hydrophonePos)

getHydrophonePos

public java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> getHydrophonePos()

getJumps

public java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> getJumps()

getChiJumps

public java.util.ArrayList<java.util.ArrayList<java.lang.Double>> getChiJumps()

getResultsAllChains

public java.util.ArrayList<MCMCTDResults> getResultsAllChains()

getResults

public java.util.ArrayList<MCMCTDResults> getResults()

changeSettings

public java.lang.Boolean changeSettings()

Description copied from interface: TimeDelayLocaliserModel

Change any settings in the algorithm.

Specified by:

changeSettings in interface TimeDelayLocaliserModel

Returns:

getSettings

public MCMCParams getSettings()

setSampleRate

public void setSampleRate(float sampleRate)

setSoundSpeed

public void setSoundSpeed(double speedOfSound)

compressMCMCResults

public static java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> compressMCMCResults(java.util.ArrayList<java.util.ArrayList<javax.vecmath.Point3f>> jumps,
                                                                                  int compressSize)

Reduces the number of MCMC jumps in order to reduce memory usage when displaying probability distributions in Java3D or other visualise components.

Parameters:

jumps - ArrayList of chain jumps to compress

compressSize- - the ideal number of points in the new array. Note: unless compressSize exactly divides the jump size then final arrays will be slightly larger than compressSize

Returns:

stop

public void stop()

Description copied from interface: TimeDelayLocaliserModel

Stop the localiser. For MCMC, which can takes minutes to compute this is important. For other localisers, such as Simplex, this is unimportant and can be left blank.

Specified by:

stop in interface TimeDelayLocaliserModel

clearResults

public void clearResults()

Clear the results from memory.