Optimizers

Class ConjugateGradient

java.lang.Object

Optimizers.ConjugateGradient

All Implemented Interfaces:

Optimizer

public class ConjugateGradient
extends java.lang.Object
implements Optimizer

Implements the Conjugate Gradient method of parameter optimisation.

This is heavily based on the BEAST implementation

Progress information can optionally be printed to either the screen or a file.

Version:

2.0

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	ConjugateGradient.ProgressLevel Enumeration of the different levels of output
static class	ConjugateGradient.Update Enumeration of the different update methods

Nested classes/interfaces inherited from interface Optimizers.Optimizer

Optimizer.OptimizerException

Constructor Summary

Constructors

Constructor and Description
ConjugateGradient() Constructor that uses default values for rigour (1e-7), update method (BEALE_SORENSON_HESTENES_STIEFEL) and output (none)
ConjugateGradient(ConjugateGradient.ProgressLevel progressLevel) Constructor that uses the given output level but default values for rigour (1e-7) and update method (BEALE_SORENSON_HESTENES_STIEFEL)
ConjugateGradient(ConjugateGradient.Update update) Constructor that uses the givemn update method and default values for rigour (1e-7) and output (none)
ConjugateGradient(ConjugateGradient.Update update, ConjugateGradient.ProgressLevel progressLevel) Constructor that uses the given update method (BEALE_SORENSON_HESTENES_STIEFEL) and output (none) but the default values for rigour (1e-7),
ConjugateGradient(double tol) Constructor that uses the given value for rigour and default values for update method (BEALE_SORENSON_HESTENES_STIEFEL) and output (none)
ConjugateGradient(double tol, ConjugateGradient.ProgressLevel progressLevel) Constructor that uses the given tolerance and output level but the default update method (BEALE_SORENSON_HESTENES_STIEFEL)
ConjugateGradient(double tol, ConjugateGradient.Update update) Constructor that uses the given tolerance and update method but the default output (none)
ConjugateGradient(double tol, ConjugateGradient.Update update, ConjugateGradient.ProgressLevel progressLevel) Constructor that uses the given values for rigour, update method and output

Method Summary

Methods

Modifier and Type	Method and Description
<R extends Likelihood> R	maximise(Optimizable<R> c, Parameters p) Maximises the likelihood, logging to screen.
<R extends Likelihood> R	maximise(Optimizable<R> c, Parameters p, java.io.File log) Maximises the likelihood, logging to a file.
<R extends Likelihood> R	restart(Optimizable<R> l, java.io.File checkPoint) Maximises the likelihood starting from a checkpoint file (see Optimizer.setCheckPointFile(java.io.File), logging to the screen.
<R extends Likelihood> R	restart(Optimizable<R> l, java.io.File checkPoint, java.io.File log) Maximises the likelihood starting from a checkpoint file (see Optimizer.setCheckPointFile(java.io.File), logging to a file.
void	setCheckPointFile(java.io.File checkPoint) Sets a checkpoint file.
void	setCheckPointFrequency(int num, java.util.concurrent.TimeUnit unit) Sets how often a checkpoint is saved (if a file has been set).
void	setMaximumRunTime(int num, java.util.concurrent.TimeUnit unit) Set a maximum time the optimiser should run for

Methods inherited from class java.lang.Object

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

Constructor Detail

ConjugateGradient

public ConjugateGradient()

Constructor that uses default values for rigour (1e-7), update method (BEALE_SORENSON_HESTENES_STIEFEL) and output (none)

ConjugateGradient

public ConjugateGradient(double tol)

Constructor that uses the given value for rigour and default values for update method (BEALE_SORENSON_HESTENES_STIEFEL) and output (none)

Parameters:

tol - The toelrance to use

ConjugateGradient

public ConjugateGradient(ConjugateGradient.Update update)

Constructor that uses the givemn update method and default values for rigour (1e-7) and output (none)

Parameters:

update - The update method to use

ConjugateGradient

public ConjugateGradient(double tol,
                 ConjugateGradient.Update update)

Constructor that uses the given tolerance and update method but the default output (none)

Parameters:

tol - The tolerance to use

update - The update method to use

ConjugateGradient

public ConjugateGradient(ConjugateGradient.ProgressLevel progressLevel)

Constructor that uses the given output level but default values for rigour (1e-7) and update method (BEALE_SORENSON_HESTENES_STIEFEL)

Parameters:

progressLevel - The output level to use

ConjugateGradient

public ConjugateGradient(double tol,
                 ConjugateGradient.ProgressLevel progressLevel)

Constructor that uses the given tolerance and output level but the default update method (BEALE_SORENSON_HESTENES_STIEFEL)

Parameters:

tol - The tolerance to use

progressLevel - The output level to use

ConjugateGradient

public ConjugateGradient(ConjugateGradient.Update update,
                 ConjugateGradient.ProgressLevel progressLevel)

Constructor that uses the given update method (BEALE_SORENSON_HESTENES_STIEFEL) and output (none) but the default values for rigour (1e-7),

Parameters:

update - the update method

progressLevel - The output level to use

ConjugateGradient

public ConjugateGradient(double tol,
                 ConjugateGradient.Update update,
                 ConjugateGradient.ProgressLevel progressLevel)

Constructor that uses the given values for rigour, update method and output

Parameters:

tol - The tolerance to use

update - The update method to use

progressLevel - The output level to use

Method Detail

maximise

public <R extends Likelihood> R maximise(Optimizable<R> c,
                                Parameters p)
                              throws GeneralException

Description copied from interface: Optimizer

Maximises the likelihood, logging to screen. Logging level should be set in the constructor of implementing classes.

Specified by:

maximise in interface Optimizer

Type Parameters:

R - The type returned by the calculator

Parameters:

c - The likelihood calculator

p - The parameters to maximise. Parameters are modified.

Returns:

The maximised likelihood (in a structure that includes most intermediate likelihoods).

Throws:

GeneralException - When there is a problem in finding an optimisable solution.

maximise

public <R extends Likelihood> R maximise(Optimizable<R> c,
                                Parameters p,
                                java.io.File log)
                              throws GeneralException

Description copied from interface: Optimizer

Maximises the likelihood, logging to a file. Logging level should be set in the constructor of implementing classes.

Specified by:

maximise in interface Optimizer

Type Parameters:

R - The type returned by the calculator

Parameters:

c - The likelihood calculator

p - The parameters to maximise. Parameters are modified.

log - The log file

Returns:

The maximised likelihood (in a structure that includes most intermediate likelihoods).

Throws:

GeneralException - When there is a problem in finding an optimisable solution

restart

public <R extends Likelihood> R restart(Optimizable<R> l,
                               java.io.File checkPoint)
                             throws GeneralException

Description copied from interface: Optimizer

Maximises the likelihood starting from a checkpoint file (see Optimizer.setCheckPointFile(java.io.File), logging to the screen. Logging level should be set in the constructor of implementing classes.

Specified by:

restart in interface Optimizer

Type Parameters:

R - The type returned by the calculator

Parameters:

l - The likelihood calculator

checkPoint - The checkpoint file

Returns:

The maximised likelihood (in a structure that includes most intermediate likelihoods).

Throws:

GeneralException - When there is a problem in finding an optimisable solution

restart

public <R extends Likelihood> R restart(Optimizable<R> l,
                               java.io.File checkPoint,
                               java.io.File log)
                             throws GeneralException

Description copied from interface: Optimizer

Maximises the likelihood starting from a checkpoint file (see Optimizer.setCheckPointFile(java.io.File), logging to a file. Logging level should be set in the constructor of implementing classes.

Specified by:

restart in interface Optimizer

Type Parameters:

R - The type returned by the calculator

Parameters:

l - The likelihood calculator

checkPoint - The checkpoint file

log - The log file

Returns:

The maximised likelihood (in a structure that includes most intermediate likelihoods).

Throws:

GeneralException - When there is a problem in finding an optimisable solution

setCheckPointFile

public void setCheckPointFile(java.io.File checkPoint)
                       throws Optimizer.OptimizerException

Description copied from interface: Optimizer

Sets a checkpoint file. If set will write a checkpoint file of the optimisers state at regular intervals (set by Optimizer.setCheckPointFrequency(int, java.util.concurrent.TimeUnit)). This checkpoint file can then be used to restart the optimiser from this state. Can be useful if optimisation is likely to take a long and the process could be stopped for some reason.

Specified by:

setCheckPointFile in interface Optimizer

Parameters:

checkPoint - The chekpoint file.

Throws:

Optimizer.OptimizerException - Thrown if the optomiser does not implement checkpoints.

setCheckPointFrequency

public void setCheckPointFrequency(int num,
                          java.util.concurrent.TimeUnit unit)
                            throws Optimizer.OptimizerException

Description copied from interface: Optimizer

Sets how often a checkpoint is saved (if a file has been set).

Specified by:

setCheckPointFrequency in interface Optimizer

Parameters:

num - The number of time units that should pass between checkpoint writes.

unit - The time unit

Throws:

Optimizer.OptimizerException - Thrown if the optomiser does not implement checkpoints.

setMaximumRunTime

public void setMaximumRunTime(int num,
                     java.util.concurrent.TimeUnit unit)
                       throws Optimizer.OptimizerException

Description copied from interface: Optimizer

Set a maximum time the optimiser should run for

Specified by:

setMaximumRunTime in interface Optimizer

Parameters:

num - The number of time units that should pass between checkpoint writes.

unit - The time unit

Throws:

Optimizer.OptimizerException - Thrown if the optimiser does not implement a maximum run time.