de.jstacs.classifiers.differentiableSequenceScoreBased
Class DiffSSBasedOptimizableFunction

java.lang.Object
  de.jstacs.algorithms.optimization.DifferentiableFunction
      de.jstacs.classifiers.differentiableSequenceScoreBased.OptimizableFunction
          de.jstacs.classifiers.differentiableSequenceScoreBased.AbstractOptimizableFunction
              de.jstacs.classifiers.differentiableSequenceScoreBased.AbstractMultiThreadedOptimizableFunction
                  de.jstacs.classifiers.differentiableSequenceScoreBased.DiffSSBasedOptimizableFunction

All Implemented Interfaces:

Function

Direct Known Subclasses:

LogGenDisMixFunction

public abstract class DiffSSBasedOptimizableFunction
extends AbstractMultiThreadedOptimizableFunction

This abstract class is the basis of all multi-threaded OptimizableFunctions that are based on DifferentiableSequenceScores.

Author:

Jens Keilwagen

Nested Class Summary

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

OptimizableFunction.KindOfParameter

Field Summary

protected DoubleList[][]	dList These DoubleLists are used during the parallel computation of the gradient.
protected IntList[][]	iList These IntLists are used during the parallel computation of the gradient.
protected LogPrior	prior The prior that is used in this function.
protected DifferentiableSequenceScore[][]	score These DifferentiableSequenceScores are used during the parallel computation.
protected int[]	shortcut These shortcuts indicate the beginning of a new part in the parameter vector.

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

params

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

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

Constructor Summary

DiffSSBasedOptimizableFunction(int threads, DifferentiableSequenceScore[] score, DataSet[] data, double[][] weights, LogPrior prior, boolean norm, boolean freeParams)
Creates an instance with the underlying infrastructure.

Method Summary

void	addTermToClassParameter(int classIndex, double term) This method adds the term to the class parameter of the class with index classIndex.
double[]	getClassParams(double[] params) Returns from the complete vector of parameters those that are for the classes.
int	getDimensionOfScope() Returns the dimension of the scope of the Function.
void	getParameters(OptimizableFunction.KindOfParameter kind, double[] erg) This method enables the user to get the parameters without creating a new array.
void	reset() Resets the all objects and pre-computed values.
abstract void	reset(DifferentiableSequenceScore[] funs) This method allows to reset the internally used functions and the corresponding objects.
protected void	setParams(int index) This method sets the parameters for thread index
protected void	setThreadIndependentParameters() This method allows to set thread independent parameters.

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

evaluateFunction, evaluateFunction, evaluateGradientOfFunction, evaluateGradientOfFunction, getNumberOfAvailableProcessors, getNumberOfThreads, joinFunction, joinGradients, setDataAndWeights, setParams, stopThreads

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

getData, getParameters, getSequenceWeights

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

Field Detail

shortcut

protected int[] shortcut

These shortcuts indicate the beginning of a new part in the parameter vector. Has to be set by reset().

score

protected DifferentiableSequenceScore[][] score

These DifferentiableSequenceScores are used during the parallel computation. score[t] contains all DifferentiableSequenceScores that are used by thread t.

dList

protected DoubleList[][] dList

These DoubleLists are used during the parallel computation of the gradient. dList[t] contains all DoubleLists that are used by thread t.

iList

protected IntList[][] iList

These IntLists are used during the parallel computation of the gradient. iList[t] contains all IntLists that are used by thread t.

prior

protected LogPrior prior

The prior that is used in this function.

Constructor Detail

DiffSSBasedOptimizableFunction

public DiffSSBasedOptimizableFunction(int threads,
                                      DifferentiableSequenceScore[] score,
                                      DataSet[] data,
                                      double[][] weights,
                                      LogPrior prior,
                                      boolean norm,
                                      boolean freeParams)
                               throws IllegalArgumentException

Creates an instance with the underlying infrastructure. Before using this instance, one has to invoke reset().

Parameters:

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

score - an array containing the DifferentiableSequenceScores that are used for determining the sequences scores

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

prior - the prior that is used for learning the parameters

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 number of classes or the dimension of the weights is not correct

Method Detail

getClassParams

public final double[] getClassParams(double[] params)

Returns from the complete vector of parameters those that are for the classes.

Parameters:

params - the current parameters

Returns:

the parameters for the classes

getDimensionOfScope

public final int getDimensionOfScope()

Description copied from interface: Function

Returns the dimension of the scope of the Function.

Returns:

the dimension of the scope: n with

setThreadIndependentParameters

protected void setThreadIndependentParameters()
                                       throws DimensionException

Description copied from class: AbstractMultiThreadedOptimizableFunction

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

Specified by:

setThreadIndependentParameters in class AbstractMultiThreadedOptimizableFunction

Throws:

DimensionException - if the dimension of the parameters is wrong

See Also:

AbstractMultiThreadedOptimizableFunction.setParams(double[])

getParameters

public void getParameters(OptimizableFunction.KindOfParameter kind,
                          double[] erg)
                   throws Exception

Description copied from class: AbstractOptimizableFunction

This method enables the user to get the parameters without creating a new array.

Specified by:

getParameters in class AbstractOptimizableFunction

Parameters:

kind - the kind of the class parameters to be returned in erg

erg - the array for the start parameters

Throws:

Exception - if the array is null or does not have the correct length

See Also:

OptimizableFunction.getParameters(KindOfParameter)

setParams

protected void setParams(int index)
                  throws DimensionException

Description copied from class: AbstractMultiThreadedOptimizableFunction

This method sets the parameters for thread index

Specified by:

setParams in class AbstractMultiThreadedOptimizableFunction

Parameters:

index - the index of the thread

Throws:

DimensionException - if the parameters could not be set

See Also:

AbstractMultiThreadedOptimizableFunction.params

addTermToClassParameter

public final void addTermToClassParameter(int classIndex,
                                          double term)

This method adds the term to the class parameter of the class with index classIndex.

Parameters:

classIndex - the index of the class

term - the term to be added to the class parameter

reset

public abstract void reset(DifferentiableSequenceScore[] funs)
                    throws Exception

This method allows to reset the internally used functions and the corresponding objects.

Parameters:

funs - the new instances

Throws:

Exception - if something went wrong

reset

public final void reset()
                 throws Exception

Description copied from class: OptimizableFunction

Resets the all objects and pre-computed values.

Specified by:

reset in class OptimizableFunction

Throws:

Exception - if something went wrong