de.jstacs.sequenceScores.statisticalModels.trainable.hmm.models

Class SamplingHigherOrderHMM

java.lang.Object

de.jstacs.sequenceScores.statisticalModels.trainable.AbstractTrainableStatisticalModel

de.jstacs.sequenceScores.statisticalModels.trainable.hmm.AbstractHMM

de.jstacs.sequenceScores.statisticalModels.trainable.hmm.models.HigherOrderHMM

de.jstacs.sequenceScores.statisticalModels.trainable.hmm.models.SamplingHigherOrderHMM

All Implemented Interfaces:

SequenceScore, StatisticalModel, TrainableStatisticalModel, Storable, Cloneable

Direct Known Subclasses:

SamplingPhyloHMM

public class SamplingHigherOrderHMM
extends HigherOrderHMM

Author:

Michael Scharfe, Jens Keilwagen

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	SamplingHigherOrderHMM.ViterbiComputation Emumeration of all possible Viterbi-Path methods

Nested classes/interfaces inherited from class de.jstacs.sequenceScores.statisticalModels.trainable.hmm.models.HigherOrderHMM

HigherOrderHMM.Type

Field Summary

Fields

Modifier and Type	Field and Description
protected BurnInTest	burnInTest This variable holds the BurnInTest used for training the model
protected boolean	hasSampled This boolean indicates if the parameters for the model were sampled

Fields inherited from class de.jstacs.sequenceScores.statisticalModels.trainable.hmm.models.HigherOrderHMM

backwardIntermediate, container, logEmission, numberOfSummands, skipInit, stateList

Fields inherited from class de.jstacs.sequenceScores.statisticalModels.trainable.hmm.AbstractHMM

bwdMatrix, emission, emissionIdx, finalState, forward, fwdMatrix, name, sostream, START_NODE, states, threads, trainingParameter, transition

Fields inherited from class de.jstacs.sequenceScores.statisticalModels.trainable.AbstractTrainableStatisticalModel

alphabets, length

Constructor Summary

Constructors

Constructor and Description
SamplingHigherOrderHMM(SamplingHMMTrainingParameterSet trainingParameterSet, String[] name, int[] emissionIdx, boolean[] forward, SamplingEmission[] emission, TransitionElement... te) This is the main constructor.
SamplingHigherOrderHMM(StringBuffer xml) The standard constructor for the interface Storable.

Method Summary

Methods

Modifier and Type	Method and Description
protected void	acceptParameters() This method can be used to accept the current parameters (and save them into a file)
protected void	appendFurtherInformation(StringBuffer xml) This method appends further information to the XML representation.
SamplingHigherOrderHMM	clone() Follows the conventions of Object's clone()-method.
protected void	createStates() This method creates states for the internal usage.
protected void	drawFromStatistics() This method draws all parameters for the current statistics
protected void	extractFurtherInformation(StringBuffer xml) This method extracts further information from the XML representation.
protected void	furtherInits(DataSet data, double[] weights) This method allows the implementation of further initializations
String	getInstanceName() Should return a short instance name such as iMM(0), BN(2), ...
protected double	getLogPosteriorFromStatistic() This method calculates the a posteriori probability for the current statistics
double	getLogProbForPath(IntList path, int startPos, Sequence seq)
double[][]	getLogStatePosteriorMatrixFor(int startPos, int endPos, Sequence seq) This method returns the log state posterior of all states for a sequence.
protected void	getNewParameters() This method set all parameters for the next sampling step
Pair<IntList,Double>	getViterbiPath(int startPos, int endPos, Sequence seq, SamplingHigherOrderHMM.ViterbiComputation compute) This method returns a viterbi path that is the optimum for the choosen ViterbiComputation method
Pair<IntList,Double>	getViterbiPathFor(int startPos, int endPos, Sequence seq)
protected String	getXMLTag() Returns the tag for the XML representation.
protected double	gibbsSampling(int startPos, int endPos, double weight, Sequence seq) This method implements a sampling step in the sampling procedure
protected void	gibbsSamplingStep(int sampling, int steps, boolean append, DataSet data, double[] weights) This method implements the next step(s) in the sampling procedure
protected void	initTraining(DataSet data, double[] weights) This methods initialize the training procedure with the given training data
boolean	isInitialized() This method can be used to determine whether the instance is initialized.
protected double	logProb(int startpos, int endpos, Sequence sequence) This method computes the logarithm of the probability of the corresponding subsequences.
protected boolean	parseNextParameterSet() This method parse a parameter set stored in file during sampling
protected boolean	parseParameterSet(int sampling, int idx) This method allows the user to parse the set of parameters with index idx of a certain sampling (from a file).
void	train(DataSet data, double[] weights) Trains the TrainableStatisticalModel object given the data as DataSet using the specified weights.

Methods inherited from class de.jstacs.sequenceScores.statisticalModels.trainable.hmm.models.HigherOrderHMM

baumWelch, createHelperVariables, estimateFromStatistics, fillBwdMatrix, fillBwdOrViterbiMatrix, fillFwdMatrix, fillLogStatePosteriorMatrix, finalize, getCharacteristics, getEmissionIndexes, getEmissions, getLogPriorTerm, getLogScoreFor, getLogScoreFor, getMaximalMarkovOrder, getNames, getNumericalCharacteristics, getTrainingParams, getTransisionElements, initialize, initializeRandomly, resetStatistics, samplePath, setSkiptInit, viterbi

Methods inherited from class de.jstacs.sequenceScores.statisticalModels.trainable.hmm.AbstractHMM

createMatrixForStatePosterior, decodePath, decodeStatePosterior, determineFinalStates, fromXML, getFinalStatePosterioriMatrix, getGraphvizRepresentation, getGraphvizRepresentation, getGraphvizRepresentation, getGraphvizRepresentation, getLogProbFor, getLogStatePosteriorMatrixFor, getNumberOfStates, getNumberOfThreads, getRunTimeException, getStatePosteriorMatrixFor, getStatePosteriorMatrixFor, getViterbiPathFor, getViterbiPathsFor, initTransition, provideMatrix, setOutputStream, toString, toXML, train

Methods inherited from class de.jstacs.sequenceScores.statisticalModels.trainable.AbstractTrainableStatisticalModel

check, emitDataSet, getAlphabetContainer, getLength, getLogProbFor, getLogProbFor, getLogScoreFor, getLogScoreFor, getLogScoreFor, toString

Methods inherited from class java.lang.Object

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

Field Detail

burnInTest

protected BurnInTest burnInTest

This variable holds the BurnInTest used for training the model

hasSampled

protected boolean hasSampled

This boolean indicates if the parameters for the model were sampled

Constructor Detail

SamplingHigherOrderHMM

public SamplingHigherOrderHMM(SamplingHMMTrainingParameterSet trainingParameterSet,
                      String[] name,
                      int[] emissionIdx,
                      boolean[] forward,
                      SamplingEmission[] emission,
                      TransitionElement... te)
                       throws Exception

This is the main constructor.

Parameters:

trainingParameterSet - the ParameterSet that determines the training algorithm and contains the necessary Parameters

name - the names of the states

emissionIdx - the indices of the emissions that should be used for each state

forward - a boolean array that indicates whether the symbol on the forward or the reverse complementary strand should be used

emission - the emissions

te - the TransitionElements building a transition

Throws:

Exception - if

• some component could not be cloned
• some the length of name, emissionIdx, or forward is not equal to the number of states
• not all emissions use the same AlphabetContainer
• the states can not be handled by the transition

SamplingHigherOrderHMM

public SamplingHigherOrderHMM(StringBuffer xml)
                       throws NonParsableException

The standard constructor for the interface Storable. Constructs an SamplingHigherOrderHMM out of an XML representation.

Parameters:

xml - the XML representation as StringBuffer

Throws:

NonParsableException - if the SamplingHigherOrderHMM could not be reconstructed out of the StringBuffer xml

Method Detail

clone

public SamplingHigherOrderHMM clone()
                             throws CloneNotSupportedException

Description copied from class: AbstractTrainableStatisticalModel

Follows the conventions of Object's clone()-method.

Specified by:

clone in interface SequenceScore

Specified by:

clone in interface TrainableStatisticalModel

Overrides:

clone in class HigherOrderHMM

Returns:

an object, that is a copy of the current AbstractTrainableStatisticalModel (the member-AlphabetContainer isn't deeply cloned since it is assumed to be immutable). The type of the returned object is defined by the class X directly inherited from AbstractTrainableStatisticalModel. Hence X's clone()-method should work as:
1. Object o = (X)super.clone();
2. all additional member variables of o defined by X that are not of simple data-types like int, double, ... have to be deeply copied
3. return o

Throws:

CloneNotSupportedException - if something went wrong while cloning

createStates

protected void createStates()

Description copied from class: AbstractHMM

This method creates states for the internal usage.

Overrides:

createStates in class HigherOrderHMM

acceptParameters

protected void acceptParameters()
                         throws IOException

This method can be used to accept the current parameters (and save them into a file)

Throws:

IOException - if the parameters could not be written

drawFromStatistics

protected void drawFromStatistics()
                           throws Exception

This method draws all parameters for the current statistics

Throws:

Exception - if the parameters could not be drawn

gibbsSampling

protected double gibbsSampling(int startPos,
                   int endPos,
                   double weight,
                   Sequence seq)
                        throws Exception

This method implements a sampling step in the sampling procedure

Parameters:

startPos - the start position in the sequence

endPos - the end position in the sequence

weight - the weight for the sequence

seq - the sequence

Returns:

the score for the sampi

Throws:

Exception - if the sampling step did not succeed

gibbsSamplingStep

protected void gibbsSamplingStep(int sampling,
                     int steps,
                     boolean append,
                     DataSet data,
                     double[] weights)
                          throws Exception

This method implements the next step(s) in the sampling procedure

Parameters:

sampling - the index of the sampling

steps - the number of sampling that should be executed

append - whether to append the sampled parameters to an existing file or to overwrite the file

data - the data used for sampling

weights - the weight for each sequence

Throws:

Exception - if something wents wrong

getNewParameters

protected void getNewParameters()
                         throws Exception

This method set all parameters for the next sampling step

Throws:

Exception - if something went wrong

train

public void train(DataSet data,
         double[] weights)
           throws Exception

Description copied from interface: TrainableStatisticalModel

Trains the TrainableStatisticalModel object given the data as DataSet using the specified weights. The weight at position i belongs to the element at position i. So the array weight should have the number of sequences in the data set as dimension. (Optionally it is possible to use weight == null if all weights have the value one.)
This method should work non-incrementally. That means the result of the following series: train(data1); train(data2) should be a fully trained model over data2 and not over data1+data2. All parameters of the model were given by the call of the constructor.

Specified by:

train in interface TrainableStatisticalModel

Overrides:

train in class HigherOrderHMM

Parameters:

data - the given sequences as DataSet

weights - the weights of the elements, each weight should be non-negative

Throws:

Exception - if the training did not succeed (e.g. the dimension of weights and the number of sequences in the data set do not match)

See Also:

DataSet.getElementAt(int), DataSet.ElementEnumerator

getInstanceName

public String getInstanceName()

Description copied from interface: SequenceScore

Should return a short instance name such as iMM(0), BN(2), ...

Specified by:

getInstanceName in interface SequenceScore

Overrides:

getInstanceName in class HigherOrderHMM

Returns:

a short instance name

isInitialized

public boolean isInitialized()

Description copied from interface: SequenceScore

This method can be used to determine whether the instance is initialized. If the instance is initialized you should be able to invoke SequenceScore.getLogScoreFor(Sequence).

Specified by:

isInitialized in interface SequenceScore

Overrides:

isInitialized in class HigherOrderHMM

Returns:

true if the instance is initialized, false otherwise

logProb

protected double logProb(int startpos,
             int endpos,
             Sequence sequence)
                  throws Exception

Description copied from class: AbstractHMM

This method computes the logarithm of the probability of the corresponding subsequences. The method does not check the AlphabetContainer and possible further features before starting the computation.

Overrides:

logProb in class AbstractHMM

Parameters:

startpos - the start position (inclusive)

endpos - the end position (inclusive)

sequence - the Sequence(s)

Returns:

the logarithm of the probability

Throws:

Exception - if the model has no parameters (for instance if it is not trained)

parseParameterSet

protected boolean parseParameterSet(int sampling,
                        int idx)
                             throws Exception

This method allows the user to parse the set of parameters with index idx of a certain sampling (from a file). The internal numbering should start with 0.

Parameters:

sampling - the index of the sampling

idx - the index of the parameter set

Returns:

true if the parameter set could be parsed

Throws:

Exception - if there is a problem with parsing the parameters

parseNextParameterSet

protected boolean parseNextParameterSet()
                                 throws Exception

This method parse a parameter set stored in file during sampling

Returns:

true if parsing was successful

Throws:

Exception - if the parameters could not be parsed

getXMLTag

protected String getXMLTag()

Description copied from class: AbstractHMM

Returns the tag for the XML representation.

Overrides:

getXMLTag in class HigherOrderHMM

Returns:

the tag for the XML representation

See Also:

AbstractHMM.fromXML(StringBuffer), AbstractHMM.toXML()

appendFurtherInformation

protected void appendFurtherInformation(StringBuffer xml)

Description copied from class: AbstractHMM

This method appends further information to the XML representation. It allows subclasses to save further parameters that are not defined in the superclass.

Overrides:

appendFurtherInformation in class HigherOrderHMM

Parameters:

xml - the XML representation

extractFurtherInformation

protected void extractFurtherInformation(StringBuffer xml)
                                  throws NonParsableException

Description copied from class: HigherOrderHMM

This method extracts further information from the XML representation. It allows subclasses to cast further parameters that are not defined in the superclass.

Overrides:

extractFurtherInformation in class HigherOrderHMM

Parameters:

xml - the XML representation

Throws:

NonParsableException - if the information could not be reconstructed out of the StringBuffer xml

initTraining

protected void initTraining(DataSet data,
                double[] weights)
                     throws Exception

This methods initialize the training procedure with the given training data

Parameters:

data - the data set used for training

weights - the weight for each sequence

Throws:

Exception - if the transition or emissions could not be initialized

furtherInits

protected void furtherInits(DataSet data,
                double[] weights)
                     throws Exception

This method allows the implementation of further initializations

Parameters:

data - the current data set

weights - the weight for each sequence

Throws:

Exception - if the init steps did not succeed

getLogStatePosteriorMatrixFor

public double[][] getLogStatePosteriorMatrixFor(int startPos,
                                       int endPos,
                                       Sequence seq)
                                         throws Exception

Description copied from class: AbstractHMM

This method returns the log state posterior of all states for a sequence.

Overrides:

getLogStatePosteriorMatrixFor in class AbstractHMM

Parameters:

startPos - the start position within the sequence

endPos - the end position within the sequence

seq - the sequence

Returns:

the score for each state an each sequence position

Throws:

Exception - if the state posterior could not be computed, for instance if the model is not trained, ...

getLogProbForPath

public double getLogProbForPath(IntList path,
                       int startPos,
                       Sequence seq)
                         throws Exception

Overrides:

getLogProbForPath in class HigherOrderHMM

Parameters:

path - the given state path

startPos - the start position within the sequence(s) (inclusive)

seq - the sequence(s)

Returns:

the logarithm of the probability for the given path and the given sequence(s)

Throws:

Exception - if the probability for the sequence given path could not be computed, for instance if the model is not trained, ...

getViterbiPathFor

public Pair<IntList,Double> getViterbiPathFor(int startPos,
                                     int endPos,
                                     Sequence seq)
                                       throws Exception

Overrides:

getViterbiPathFor in class HigherOrderHMM

Parameters:

startPos - the start position within the sequence

endPos - the end position within the sequence

seq - the sequence

Returns:

a Pair containing the viterbi state path and the corresponding score

Throws:

Exception - if the viterbi path could not be computed, for instance if the model is not trained, ...

getViterbiPath

public Pair<IntList,Double> getViterbiPath(int startPos,
                                  int endPos,
                                  Sequence seq,
                                  SamplingHigherOrderHMM.ViterbiComputation compute)
                                    throws Exception

This method returns a viterbi path that is the optimum for the choosen ViterbiComputation method

Parameters:

startPos - the start position in the sequence

endPos - the end position in the sequence

seq - the sequence

compute - the ViterbiComputation method

Returns:

the pair of path and score

Throws:

Exception - if the parameters could not be parsed from file

getLogPosteriorFromStatistic

protected double getLogPosteriorFromStatistic()

This method calculates the a posteriori probability for the current statistics

Returns:

the logarithm of the a posteriori probability for the current statistics