de.jstacs.classifiers.differentiableSequenceScoreBased

Class AbstractMultiThreadedOptimizableFunction

java.lang.Object

de.jstacs.algorithms.optimization.DifferentiableFunction

de.jstacs.classifiers.differentiableSequenceScoreBased.OptimizableFunction

de.jstacs.classifiers.differentiableSequenceScoreBased.AbstractOptimizableFunction

de.jstacs.classifiers.differentiableSequenceScoreBased.AbstractMultiThreadedOptimizableFunction

All Implemented Interfaces:

Function, MultiThreadedFunction

Direct Known Subclasses:

DiffSSBasedOptimizableFunction

public abstract class AbstractMultiThreadedOptimizableFunction
extends AbstractOptimizableFunction
implements MultiThreadedFunction

This class enables the user to exploit all CPUs of an computer by using threads. Each thread computes the function for a part of the data. The number of compute threads can be determined in the constructor.

The main goal is on the one hand to hide the complete infrastructure that is used for the multi-threading from the programmer and on the other hand to provide an simple interface in form of few abstract methods.

It is very important for this class that the clone() method (of the used items) works correctly, since each thread works on its own clones.

Author:

Jens Keilwagen

Nested Class Summary

Nested classes/interfaces inherited from class de.jstacs.classifiers.differentiableSequenceScoreBased.OptimizableFunction

OptimizableFunction.KindOfParameter

Field Summary

Fields

Modifier and Type	Field and Description
protected double[]	params This is a pointer for the current parameters.
protected de.jstacs.classifiers.differentiableSequenceScoreBased.AbstractMultiThreadedOptimizableFunction.Worker[]	worker The worker threads for computing value and gradient of this AbstractMultiThreadedOptimizableFunction

Fields inherited from class de.jstacs.classifiers.differentiableSequenceScoreBased.AbstractOptimizableFunction

cl, clazz, data, freeParams, logClazz, norm, sum, weights

Constructor Summary

Constructors

Constructor and Description
AbstractMultiThreadedOptimizableFunction(int threads, DataSet[] data, double[][] weights, boolean norm, boolean freeParams) The constructor for an multi-threaded instance.

Method Summary

Modifier and Type	Method and Description
double	evaluateFunction(double[] x) Evaluates the function at a certain vector (in mathematical sense) x.
protected abstract void	evaluateFunction(int index, int startClass, int startSeq, int endClass, int endSeq) This method evaluates the function for a part of the data.
double[]	evaluateGradientOfFunction(double[] x) Evaluates the gradient of a function at a certain vector (in mathematical sense) x, i.e., .
protected abstract void	evaluateGradientOfFunction(int index, int startClass, int startSeq, int endClass, int endSeq) This method evaluates the gradient of the function for a part of the data.
static int	getNumberOfAvailableProcessors() This method returns the number of available processors.
int	getNumberOfThreads() Returns the number of used threads for evaluating the function and for determining the gradient of the function.
protected abstract double	joinFunction() This method joins the partial results that have been computed using evaluateFunction(int, int, int, int, int).
protected abstract double[]	joinGradients() This method joins the gradients of each part that have been computed using evaluateGradientOfFunction(int, int, int, int, int).
protected void	prepareThreads() Assigns parts of the data to the threads
void	setDataAndWeights(DataSet[] data, double[][] weights) This method sets the data set and the sequence weights to be used.
void	setParams(double[] params) Sets the current values as parameters.
protected abstract void	setParams(int index) This method sets the parameters for thread index
protected abstract void	setThreadIndependentParameters() This method allows to set thread independent parameters.
void	stopThreads() This method can and should be used to stop all threads if they are not needed any longer.

Methods inherited from class de.jstacs.classifiers.differentiableSequenceScoreBased.AbstractOptimizableFunction

getData, getParameters, getParameters, getSequenceWeights

Methods inherited from class de.jstacs.classifiers.differentiableSequenceScoreBased.OptimizableFunction

reset

Methods inherited from class de.jstacs.algorithms.optimization.DifferentiableFunction

findOneDimensionalMin

Methods inherited from class java.lang.Object

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

Methods inherited from interface de.jstacs.algorithms.optimization.Function

getDimensionOfScope

Field Detail

worker

protected de.jstacs.classifiers.differentiableSequenceScoreBased.AbstractMultiThreadedOptimizableFunction.Worker[] worker

The worker threads for computing value and gradient of this AbstractMultiThreadedOptimizableFunction

params

protected double[] params

This is a pointer for the current parameters.

Constructor Detail

AbstractMultiThreadedOptimizableFunction

public AbstractMultiThreadedOptimizableFunction(int threads,
                                                DataSet[] data,
                                                double[][] weights,
                                                boolean norm,
                                                boolean freeParams)
                                         throws IllegalArgumentException

The constructor for an multi-threaded instance.

Parameters:

threads - the number of threads used for evaluating the function and determining the gradient of the function

data - the array of DataSets containing the data that is needed to evaluate the function

weights - the weights for each Sequence in each DataSet of data

norm - the switch for using the normalization (division by the number of sequences)

freeParams - the switch for using only the free parameters

Throws:

IllegalArgumentException - if the number of threads is not positive, the DataSets contain too less Sequence for the number of threads, the number of classes or the dimension of the weights is not correct

Method Detail

getNumberOfAvailableProcessors

public static final int getNumberOfAvailableProcessors()

This method returns the number of available processors. So if no other job is running this is the recommended number of threads.

Returns:

the number of available processors

setDataAndWeights

public void setDataAndWeights(DataSet[] data,
                              double[][] weights)
                       throws IllegalArgumentException

Description copied from class: OptimizableFunction

This method sets the data set and the sequence weights to be used. It also allows to do further preparation for the computation on this data.

Overrides:

setDataAndWeights in class AbstractOptimizableFunction

Parameters:

data - the data sets

weights - the sequence weights for each sequence in each data set

Throws:

IllegalArgumentException - if the data or the weights can not be used

prepareThreads

protected void prepareThreads()

Assigns parts of the data to the threads

evaluateGradientOfFunction

public double[] evaluateGradientOfFunction(double[] x)
                                    throws DimensionException,
                                           EvaluationException

Description copied from class: DifferentiableFunction

Evaluates the gradient of a function at a certain vector (in mathematical sense) x, i.e., .

Specified by:

evaluateGradientOfFunction in class DifferentiableFunction

Parameters:

x - the current vector

Returns:

the evaluation of the gradient of a function, has dimension Function.getDimensionOfScope()

Throws:

DimensionException - if dim(x) != n, with

EvaluationException - if there was something wrong during the evaluation of the gradient

See Also:

Function.getDimensionOfScope()

evaluateGradientOfFunction

protected abstract void evaluateGradientOfFunction(int index,
                                                   int startClass,
                                                   int startSeq,
                                                   int endClass,
                                                   int endSeq)

This method evaluates the gradient of the function for a part of the data.

Parameters:

index - the index of the part

startClass - the index of the start class

startSeq - the index of the start sequence

endClass - the index of the end class (inclusive)

endSeq - the index of the end sequence (exclusive)

joinGradients

protected abstract double[] joinGradients()
                                   throws EvaluationException

This method joins the gradients of each part that have been computed using evaluateGradientOfFunction(int, int, int, int, int).

Returns:

the gradient

Throws:

EvaluationException - if the gradient could not be evaluated properly

evaluateFunction

public double evaluateFunction(double[] x)
                        throws DimensionException,
                               EvaluationException

Description copied from interface: Function

Evaluates the function at a certain vector (in mathematical sense) x.

Specified by:

evaluateFunction in interface Function

Parameters:

x - the current vector

Returns:

the evaluation of the function

Throws:

DimensionException - if dim(x) != n, with

EvaluationException - if there was something wrong during the evaluation of the function

evaluateFunction

protected abstract void evaluateFunction(int index,
                                         int startClass,
                                         int startSeq,
                                         int endClass,
                                         int endSeq)
                                  throws EvaluationException

This method evaluates the function for a part of the data.

Parameters:

index - the index of the part

startClass - the index of the start class

startSeq - the index of the start sequence

endClass - the index of the end class (inclusive)

endSeq - the index of the end sequence (exclusive)

Throws:

EvaluationException - if the gradient could not be evaluated properly

joinFunction

protected abstract double joinFunction()
                                throws EvaluationException,
                                       DimensionException

This method joins the partial results that have been computed using evaluateFunction(int, int, int, int, int).

Returns:

the value of the function

Throws:

EvaluationException - if the gradient could not be evaluated properly

DimensionException - if the parameters could not be set

setParams

public final void setParams(double[] params)
                     throws DimensionException

Description copied from class: OptimizableFunction

Sets the current values as parameters.

Specified by:

setParams in class OptimizableFunction

Parameters:

params - the current values

Throws:

DimensionException - if the dimension of the current values does not match with the dimension of the internal parameters

setThreadIndependentParameters

protected abstract void setThreadIndependentParameters()
                                                throws DimensionException

This method allows to set thread independent parameters. It also allows to check some constraints as for instance the dimension of the parameters.

Throws:

DimensionException - if the dimension of the parameters is wrong

See Also:

setParams(double[])

setParams

protected abstract void setParams(int index)
                           throws DimensionException

This method sets the parameters for thread index

Parameters:

index - the index of the thread

Throws:

DimensionException - if the parameters could not be set

See Also:

params

stopThreads

public final void stopThreads()

This method can and should be used to stop all threads if they are not needed any longer.

Specified by:

stopThreads in interface MultiThreadedFunction

getNumberOfThreads

public final int getNumberOfThreads()

Returns the number of used threads for evaluating the function and for determining the gradient of the function.

Specified by:

getNumberOfThreads in interface MultiThreadedFunction

Returns:

the number of used threads for evaluating the function and for determining the gradient of the function