Package Localiser.algorithms.genericLocaliser.MCMC

Class MCMCParams2

java.lang.Object

Localiser.algorithms.genericLocaliser.MCMC.MCMCParams2

All Implemented Interfaces:

Serializable, Cloneable, ManagedParameters

Direct Known Subclasses:

MimplexParams

public class MCMCParams2
extends Object
implements Serializable, Cloneable, ManagedParameters

Paramters for running a Marklov chain Monte Carlo algorithm.

Author:

Jamie Macaulay

See Also:

• Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
int	chainAnalysis	Parameters for analysis of chains
double[][]	chainStartDispersion	Where in parameter space the chains should start.
int	clusterAnalysis	The type of cluster analysis to use to merge the different chains if they converge on the same or ambiguous results.
boolean	cylindricalCoOrdinates	Use cylindrical jumps.
static final int	GELMAM_RUBIN_DIAGNOSIS
static final int	IGNORE_PERCENTAGE
double[]	jumpSize	The size of Gaussian distribution for the jump in each dimension;
static final int	K_MEANS	Use K_MEANS to cluster the different chains
int	kmeanAttempts	The number of times to perform a k-means algorithm on results The algorithm starts clusters at random locations.
int	kmeansIterations	The number of iterations for each k-means attempt.
double	maxClusterSize
static final int	MEDIAN
Integer	nKMeans
static final int	NONE	Do not attempt to cluster the chains.
int	numberOfChains	The number of seperate chains to use in the simulation
int	numberOfJumps	The number of jumps in each Markov chain to attempt.
double	percentageToIgnore
static final long	serialVersionUID

Constructor Summary

Constructors

Constructor	Description
MCMCParams2()	Constructor for MCMC params.

Method Summary

Modifier and Type	Method	Description
MCMCParams2	clone()	Timing error to add to time delays.
PamParameterSet	getParameterSet()	Get a set of data that describes all of the parameters in a class
double[][]	setChainDispersion(double dispersion, int nDim)	Set chain start dispersion to one value for all dimensions.
void	setJumpSize(double jumpsize, int nDim)	Set the chain jump size- sets the jump size of all dimensions to the input jump size.

Methods inherited from class java.lang.Object

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

Field Details

serialVersionUID

public static final long serialVersionUID

See Also:

• Constant Field Values

jumpSize

public double[] jumpSize

The size of Gaussian distribution for the jump in each dimension;

numberOfJumps

public int numberOfJumps

The number of jumps in each Markov chain to attempt. (not the total successful jumps that should be reached before the chain ends)

chainStartDispersion

public double[][] chainStartDispersion

Where in parameter space the chains should start. Each value in the array is for one dimension. When chains start a random number between the two numbers is chosen as the start position in that dimension for the chain.

numberOfChains

public int numberOfChains

The number of seperate chains to use in the simulation

cylindricalCoOrdinates

public boolean cylindricalCoOrdinates

Use cylindrical jumps.

chainAnalysis

public int chainAnalysis

Parameters for analysis of chains

GELMAM_RUBIN_DIAGNOSIS

public static final int GELMAM_RUBIN_DIAGNOSIS

See Also:

• Constant Field Values

IGNORE_PERCENTAGE

public static final int IGNORE_PERCENTAGE

See Also:

• Constant Field Values

MEDIAN

public static final int MEDIAN

See Also:

• Constant Field Values

percentageToIgnore

public double percentageToIgnore

clusterAnalysis

public int clusterAnalysis

The type of cluster analysis to use to merge the different chains if they converge on the same or ambiguous results. See K_MEANS and NONE constants.

NONE

public static final int NONE

Do not attempt to cluster the chains.

See Also:

• Constant Field Values

K_MEANS

public static final int K_MEANS

Use K_MEANS to cluster the different chains

See Also:

• Constant Field Values

nKMeans

public Integer nKMeans

maxClusterSize

public double maxClusterSize

kmeanAttempts

public int kmeanAttempts

The number of times to perform a k-means algorithm on results The algorithm starts clusters at random locations.

kmeansIterations

public int kmeansIterations

The number of iterations for each k-means attempt. K-means converges to a result but requires a certain number of iteration to do so.

Constructor Details

MCMCParams2

public MCMCParams2()

Constructor for MCMC params.

Method Details

setJumpSize

public void setJumpSize(double jumpsize,
 int nDim)

Set the chain jump size- sets the jump size of all dimensions to the input jump size.

Parameters:

jumpsize - the jump size for all dimensions.

nDim - the number of dimensions.

setChainDispersion

public double[][] setChainDispersion(double dispersion,
 int nDim)

Set chain start dispersion to one value for all dimensions. The start dispersion is a random number centered on zero and from a distribution with a size defined by the dispersion input parameter.

Parameters:

dispersion - the chain start dispersion for all dimensions.

clone

public MCMCParams2 clone()

Timing error to add to time delays. This is here for legacy reasons.

getParameterSet

public PamParameterSet getParameterSet()

Description copied from interface: ManagedParameters

Get a set of data that describes all of the parameters in a class

Specified by:

getParameterSet in interface ManagedParameters

Returns:

description of the parameters in a class.