Package com.irurueta.numerical

Class AccurateMaximumLikelihoodEstimator

java.lang.Object

com.irurueta.numerical.MaximumLikelihoodEstimator

com.irurueta.numerical.AccurateMaximumLikelihoodEstimator

public class AccurateMaximumLikelihoodEstimator
extends MaximumLikelihoodEstimator

Class to estimate the most likely value from a series of samples assumed to be normally distributed. This implementation will first use an internal HistogramMaximumLikelihood to estimate the most likely value using a histogram, and then it will refine the solution by using a BrentSingleOptimizer in order to find the maximum of the probability distribution function assuming that such function is computed by aggregating small Gaussians (of size gaussianSigma) centered at the location of each sample.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
private class	AccurateMaximumLikelihoodEstimator.EvaluatorListener	Internal class used by the BrentSingleOptimizer in order to evaluate the aggregation of Gaussians for all the samples in input data array with a high degree of precision.

Field Summary

Fields

Modifier and Type	Field	Description
static final boolean	DEFAULT_USE_HISTOGRAM_INITIAL_SOLUTION	Boolean indicating if an initial solution should be obtained first by using the Histogram method.
static final double	EPS	Value to be considered as the machine precision.
private HistogramMaximumLikelihoodEstimator	internalEstimator	Internal maximum likelihood estimator based on the Histogram method.
private BrentSingleOptimizer	optimizer	Internal optimizer to find the true maximum of the probability distribution function.
private boolean	useHistogramInitialSolution	Boolean that indicates that an initial coarse solution will be computed first by using an internal HistogramMaximumLikelihoodEstimator in order to initialize the internal BrentSingleOptimizer to obtain a more accurate solution.

Fields inherited from class com.irurueta.numerical.MaximumLikelihoodEstimator

areMinMaxAvailable, DEFAULT_GAUSSIAN_SIGMA, DEFAULT_METHOD, gaussianSigma, inputData, locked, maxValue, MIN_GAUSSIAN_SIGMA, minValue

Constructor Summary

Constructors

Constructor	Description
AccurateMaximumLikelihoodEstimator()	Empty constructor.
AccurateMaximumLikelihoodEstimator(double[] inputData, double gaussianSigma, boolean useHistogramInitialSolution)	Constructor
AccurateMaximumLikelihoodEstimator(double gaussianSigma, boolean useHistogramInitialSolution)	Constructor.
AccurateMaximumLikelihoodEstimator(double minValue, double maxValue, double[] inputData, double gaussianSigma, boolean useHistogramInitialSolution)	Constructor.

Method Summary

Modifier and Type	Method	Description
double	estimate()	Starts the estimation of the most likely value contained within provided input data array.
MaximumLikelihoodEstimatorMethod	getMethod()	Returns method to be used for maximum likelihood estimation, which for this class is MaximumLikelihoodEstimatorMethod.
boolean	isHistogramInitialSolutionUsed()	Returns boolean that indicates that an initial coarse solution will be computed first by using an internal HistogramMaximumLikelihoodEstimator in order to initialize the internal BrentSingleOptimizer to obtain a more accurate solution.
void	setHistogramInitialSolutionUsed(boolean used)	Sets boolean that indicates that an initial coarse solution will be computed first by using an internal HistogramMaximumLikelihoodEstimator in order to initialize the internal BrentSingleOptimizer to obtain a more accurate solution.

Methods inherited from class com.irurueta.numerical.MaximumLikelihoodEstimator

areMinMaxValuesAvailable, computeMinMaxValues, create, create, create, create, create, create, create, create, create, getGaussianSigma, getInputData, getMaxValue, getMinValue, isInputDataAvailable, isLocked, isReady, setGaussianSigma, setInputData, setInputData, setMinMaxValues

Methods inherited from class java.lang.Object

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

Field Details

DEFAULT_USE_HISTOGRAM_INITIAL_SOLUTION

public static final boolean DEFAULT_USE_HISTOGRAM_INITIAL_SOLUTION

Boolean indicating if an initial solution should be obtained first by using the Histogram method. It is suggested to always enable this option.

See Also:

• Constant Field Values

EPS

public static final double EPS

Value to be considered as the machine precision.

See Also:

• Constant Field Values

useHistogramInitialSolution

private boolean useHistogramInitialSolution

Boolean that indicates that an initial coarse solution will be computed first by using an internal HistogramMaximumLikelihoodEstimator in order to initialize the internal BrentSingleOptimizer to obtain a more accurate solution.

internalEstimator

private HistogramMaximumLikelihoodEstimator internalEstimator

Internal maximum likelihood estimator based on the Histogram method.

optimizer

private BrentSingleOptimizer optimizer

Internal optimizer to find the true maximum of the probability distribution function. Because a BrentSingleOptimizer is only guaranteed to obtain local minima/maxima, it is preferred to start the optimizer near the true solution to be found, for that reason it is suggested to always use the Histogram initial solution as coarse approximation to start the optimizer and get a more accurate solution.

Constructor Details

AccurateMaximumLikelihoodEstimator

public AccurateMaximumLikelihoodEstimator(double gaussianSigma,
 boolean useHistogramInitialSolution)

Constructor.

Parameters:

gaussianSigma - Gaussian sigma to be used on each sample.

useHistogramInitialSolution - Boolean indicating whether an internal HistogramMaximumLikelihoodEstimator will be used to obtain a coarse initial solution to initialize the BrentSingleOptimizer. It is suggested to set this value always to true.

Throws:

IllegalArgumentException - Raised if provided Gaussian sigma is negative or zero.

AccurateMaximumLikelihoodEstimator

public AccurateMaximumLikelihoodEstimator()

Empty constructor.

AccurateMaximumLikelihoodEstimator

public AccurateMaximumLikelihoodEstimator(double[] inputData,
 double gaussianSigma,
 boolean useHistogramInitialSolution)

Constructor

Parameters:

inputData - Array containing input data where most likely value must be estimated from.

gaussianSigma - Gaussian sigma to be used on each sample.

useHistogramInitialSolution - Boolean indicating whether an internal HistogramMaximumLikelihoodEstimator will be used to obtain a coarse initial solution to initialize the BrentSingleOptimizer. It is suggested to set this value always to true.

Throws:

IllegalArgumentException - Raised if provided Gaussian sigma is negative or zero.

AccurateMaximumLikelihoodEstimator

public AccurateMaximumLikelihoodEstimator(double minValue,
 double maxValue,
 double[] inputData,
 double gaussianSigma,
 boolean useHistogramInitialSolution)

Constructor.

Parameters:

minValue - Minimum value assumed to be contained within input data array.

maxValue - Maximum value assumed to be contained within input data array.

inputData - Array containing input data where most likely value must be estimated from.

gaussianSigma - Gaussian sigma to be used on each sample.

useHistogramInitialSolution - Boolean indicating whether an internal HistogramMaximumLikelihoodEstimator will be used to obtain a coarse initial solution to initialize the BrentSingleOptimizer. It is suggested to set this value always to true.

Throws:

IllegalArgumentException - Raised if provided Gaussian sigma is negative or zero, or if minValue < maxValue.

Method Details

getMethod

public MaximumLikelihoodEstimatorMethod getMethod()

Returns method to be used for maximum likelihood estimation, which for this class is MaximumLikelihoodEstimatorMethod. ACCURATE_MAXIMUM_LIKELIHOOD_ESTIMATOR.

Specified by:

getMethod in class MaximumLikelihoodEstimator

Returns:

Method for maximum likelihood estimation.

isHistogramInitialSolutionUsed

public boolean isHistogramInitialSolutionUsed()

Returns boolean that indicates that an initial coarse solution will be computed first by using an internal HistogramMaximumLikelihoodEstimator in order to initialize the internal BrentSingleOptimizer to obtain a more accurate solution.

Returns:

True if an initial coarse solution if found by the Histogram method, false otherwise.

setHistogramInitialSolutionUsed

public void setHistogramInitialSolutionUsed(boolean used)
                                     throws LockedException

Sets boolean that indicates that an initial coarse solution will be computed first by using an internal HistogramMaximumLikelihoodEstimator in order to initialize the internal BrentSingleOptimizer to obtain a more accurate solution.

Parameters:

used - True if an initial coarse solution will be found by the Histogram method, false otherwise.

Throws:

LockedException - Exception raised if this instance is locked. This method can only be executed when computations finish and this instance becomes unlocked.

estimate

public double estimate()
                throws LockedException,
NotReadyException

Starts the estimation of the most likely value contained within provided input data array.

Specified by:

estimate in class MaximumLikelihoodEstimator

Returns:

The most likely value.

Throws:

LockedException - Exception raised if this instance is locked. This method can only be executed when computations finish and this instance becomes unlocked.

NotReadyException - Exception raised if this instance is not yet ready.

See Also:

• MaximumLikelihoodEstimator.isReady()