de.jstacs.sequenceScores.statisticalModels.trainable.hmm.transitions
Class BasicHigherOrderTransition

java.lang.Object
  de.jstacs.sequenceScores.statisticalModels.trainable.hmm.transitions.BasicHigherOrderTransition

All Implemented Interfaces:

TrainableTransition, Transition, TransitionWithSufficientStatistic, Storable, Cloneable

Direct Known Subclasses:

HigherOrderTransition

public class BasicHigherOrderTransition
extends Object
implements TrainableTransition

This class implements the basic transition that allows to be trained using the viterbi or the Baum-Welch algorithm. Layer 0, context 0 encodes the start state.

Author:

Jan Grau, Jens Keilwagen

Nested Class Summary

static class

BasicHigherOrderTransition.AbstractTransitionElement This class declares the probability distribution for a given context, i.e. it contains all possible transition and the corresponding probabilities for a given set offset previously visited states.

Field Summary

protected boolean[]	isSilent A vector indicating for each state whether it is silent or not.
protected int[][][]	lookup The lookup table for spare context en- and decoding.
protected int	maximalMarkovOrder The maximal Markov order of the transition.
protected int	maxInDegree The maximal in-degree of any state.
protected BasicHigherOrderTransition.AbstractTransitionElement[]	transitions The internally used BasicHigherOrderTransition.AbstractTransitionElements.

Constructor Summary

BasicHigherOrderTransition(boolean[] isSilent, BasicHigherOrderTransition.AbstractTransitionElement... transitions)
The main constructor.

BasicHigherOrderTransition(StringBuffer xml)
The standard constructor for the interface Storable.

Method Summary

void	addToStatistic(int layer, int index, int childIdx, double weight, Sequence sequence, int sequencePosition) This method allows to add a certain weight to the sufficient statistic of a specific transition.
protected void	appendFurtherInformation(StringBuffer xml) This method appends further information to the XML representation.
BasicHigherOrderTransition	clone() This method returns a deep clone of the current instance.
void	drawParametersFromStatistic() This method allows to draw parameters from the sufficient statistic, i.e., to draw from the posterior.
void	estimateFromStatistic() This method estimates the parameter of the transition using the internal sufficient statistic.
protected void	extractFurtherInformation(StringBuffer xml) This method extracts further information from the XML representation.
void	fillTransitionInformation(int layer, int index, int childIdx, int[] container) This method fills all relevant information for a specific edge in a given container.
int	getChildIdx(int layer, int index, int state) This method returns the child index of the state, if this state is no child of the context -1 is returned
String	getGraphizNetworkRepresentation(NumberFormat nf, String arrowOption, boolean graphical) This method returns a String representation of the structure that can be used in Graphviz to create an image.
int	getLastContextState(int layer, int index) The method returns the index of the state of the context, if there is no context -1 is returned.
double	getLogGammaScoreFromStatistic() This method calculates a score for the current statistics, which is independent from the current parameters In general the gamma-score is a product of gamma-functions parameterized with the current statistics
double	getLogPriorTerm() Returns a value that is proportional to the log of the prior.
double	getLogScoreFor(int layer, int index, int childIdx, Sequence sequence, int sequencePosition) This method returns the logarithm of the score for the transition.
int	getMaximalInDegree() This method returns the maximal out degree of any context used in this transition instance.
int	getMaximalMarkovOrder() This method returns the maximal used Markov order.
int	getMaximalNumberOfChildren() This method returns the maximal number of children for any context used in this transition instance.
int	getNumberOfChildren(int layer, int index) This method returns the number of children states for given index, i.e. context, and a given layer of the matrix.
int	getNumberOfIndexes(int layer) This method computes the number of different indexes for a given layer of the matrix.
int	getNumberOfStates() This method returns the number of states underlying this transition instance.
protected int	getTransitionElementIndex(int layer, int index) This method return the index of the BasicHigherOrderTransition.AbstractTransitionElement using the lookup table.
protected String	getXMLTag() The method returns the XML tag used during saving and loading the transition.
boolean	hasAnySelfTransitions() This method answers the question whether the instance models any self transitions.
void	initializeRandomly() This method randomly initializes the parameters of the transition.
boolean[]	isAbsoring() This method returns for each state whether it is absorbing or not.
void	joinStatistics(Transition... transitions) This method joins the statistics of different instances and sets this joined statistic as statistic of each instance.
void	resetStatistic() This method reset the internal sufficient statistics that can be used for estimating the parameters.
void	setParameters(Transition t) Set values of parameters of the instance to the value of the parameters of the given instance.
String	toString()
String	toString(String[] stateNames) This method returns a String representation of the Transition using the given names of the states.
StringBuffer	toXML() This method returns an XML representation as StringBuffer of an instance of the implementing class.

Methods inherited from class java.lang.Object

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

Field Detail

transitions

protected BasicHigherOrderTransition.AbstractTransitionElement[] transitions

The internally used BasicHigherOrderTransition.AbstractTransitionElements.

isSilent

protected boolean[] isSilent

A vector indicating for each state whether it is silent or not.

maximalMarkovOrder

protected int maximalMarkovOrder

The maximal Markov order of the transition.

maxInDegree

protected int maxInDegree

The maximal in-degree of any state.

lookup

protected int[][][] lookup

The lookup table for spare context en- and decoding.

Constructor Detail

BasicHigherOrderTransition

public BasicHigherOrderTransition(boolean[] isSilent,
                                  BasicHigherOrderTransition.AbstractTransitionElement... transitions)
                           throws Exception

The main constructor.

Parameters:

transitions - the BasicHigherOrderTransition.AbstractTransitionElements for the internal use

isSilent - an array indicating for each state whether it is silent or not

Throws:

Exception - if an error occurs during checking the BasicHigherOrderTransition.AbstractTransitionElements and creating internal fields

BasicHigherOrderTransition

public BasicHigherOrderTransition(StringBuffer xml)
                           throws NonParsableException

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

Parameters:

xml - the XML representation as StringBuffer

Throws:

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

Method Detail

toXML

public StringBuffer toXML()

Description copied from interface: Storable

This method returns an XML representation as StringBuffer of an instance of the implementing class.

Specified by:

toXML in interface Storable

Returns:

the XML representation

getXMLTag

protected String getXMLTag()

The method returns the XML tag used during saving and loading the transition.

Returns:

he XML tag used during saving and loading the transition.

appendFurtherInformation

protected void appendFurtherInformation(StringBuffer xml)

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

Parameters:

xml - the XML representation

extractFurtherInformation

protected void extractFurtherInformation(StringBuffer xml)
                                  throws NonParsableException

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

Parameters:

xml - the XML representation

Throws:

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

clone

public BasicHigherOrderTransition clone()
                                 throws CloneNotSupportedException

Description copied from interface: Transition

This method returns a deep clone of the current instance.

Specified by:

clone in interface Transition

Overrides:

clone in class Object

Returns:

a deep clone of the current instance.

Throws:

CloneNotSupportedException - if the instance could not be cloned

resetStatistic

public void resetStatistic()

Description copied from interface: TransitionWithSufficientStatistic

This method reset the internal sufficient statistics that can be used for estimating the parameters.

Specified by:

resetStatistic in interface TransitionWithSufficientStatistic

joinStatistics

public void joinStatistics(Transition... transitions)

Description copied from interface: TransitionWithSufficientStatistic

This method joins the statistics of different instances and sets this joined statistic as statistic of each instance. This method might be used for instance in a multi-threaded optimization to join partial statistics.

Specified by:

joinStatistics in interface TransitionWithSufficientStatistic

Parameters:

transitions - the transitions to be joined

addToStatistic

public void addToStatistic(int layer,
                           int index,
                           int childIdx,
                           double weight,
                           Sequence sequence,
                           int sequencePosition)

Description copied from interface: TransitionWithSufficientStatistic

This method allows to add a certain weight to the sufficient statistic of a specific transition.

Specified by:

addToStatistic in interface TransitionWithSufficientStatistic

Parameters:

layer - the layer of the matrix

index - the index encoding the context

childIdx - the index of the child

weight - the weight added to the sufficient statistic

sequence - the sequence

sequencePosition - the position within the sequence

estimateFromStatistic

public void estimateFromStatistic()

Description copied from interface: TrainableTransition

This method estimates the parameter of the transition using the internal sufficient statistic.

Specified by:

estimateFromStatistic in interface TrainableTransition

drawParametersFromStatistic

public void drawParametersFromStatistic()
                                 throws Exception

This method allows to draw parameters from the sufficient statistic, i.e., to draw from the posterior. If no data has been added using addToStatistic(int, int, int, double, Sequence, int), the method draws from the prior.

Throws:

Exception - if for instance the prior has some illegal hyper-parameters (e.g. 0)

See Also:

BasicHigherOrderTransition.AbstractTransitionElement.drawParametersFromStatistic()

getMaximalMarkovOrder

public int getMaximalMarkovOrder()

Description copied from interface: Transition

This method returns the maximal used Markov order.

Specified by:

getMaximalMarkovOrder in interface Transition

Returns:

maximal used Markov order

getLogPriorTerm

public double getLogPriorTerm()

Description copied from interface: Transition

Returns a value that is proportional to the log of the prior. For maximum likelihood (ML) 0 should be returned.

Specified by:

getLogPriorTerm in interface Transition

Returns:

a value that is proportional to the log of the prior

See Also:

StatisticalModel.getLogPriorTerm()

getNumberOfStates

public int getNumberOfStates()

Description copied from interface: Transition

This method returns the number of states underlying this transition instance.

Specified by:

getNumberOfStates in interface Transition

Returns:

the number of states underlying this transition instance

initializeRandomly

public void initializeRandomly()

Description copied from interface: Transition

This method randomly initializes the parameters of the transition.

Specified by:

initializeRandomly in interface Transition

getGraphizNetworkRepresentation

public String getGraphizNetworkRepresentation(NumberFormat nf,
                                              String arrowOption,
                                              boolean graphical)

Description copied from interface: Transition

This method returns a String representation of the structure that can be used in Graphviz to create an image.

Specified by:

getGraphizNetworkRepresentation in interface Transition

Parameters:

nf - the NumberFormat used for the probabilities, if null no probabilities will we written

arrowOption - this parameter gives the possibility to set some arrow option

graphical - represent transition probabilities as thickness of edges instead of textual output

Returns:

String representation of the state

fillTransitionInformation

public void fillTransitionInformation(int layer,
                                      int index,
                                      int childIdx,
                                      int[] container)

Description copied from interface: Transition

This method fills all relevant information for a specific edge in a given container.

Specified by:

fillTransitionInformation in interface Transition

Parameters:

layer - the layer in the matrix

index - the index encoding the context

childIdx - the index of the child that will be visited

container - container[0] the index of the state; container[1] the index encoding the new context; container[2] the distance for the new layer, i.e. 0 for the same layer (=silent state), and 1 for the next layer (=non-silent state)

getTransitionElementIndex

protected final int getTransitionElementIndex(int layer,
                                              int index)

This method return the index of the BasicHigherOrderTransition.AbstractTransitionElement using the lookup table.

Parameters:

layer - the layer of the matrix

index - the index encoding the context

Returns:

the index of the BasicHigherOrderTransition.AbstractTransitionElement with respect to transitions

getNumberOfChildren

public int getNumberOfChildren(int layer,
                               int index)

Description copied from interface: Transition

This method returns the number of children states for given index, i.e. context, and a given layer of the matrix.

Specified by:

getNumberOfChildren in interface Transition

Parameters:

layer - the layer of the matrix

index - the index encoding the context

Returns:

the number of children states

getLogScoreFor

public double getLogScoreFor(int layer,
                             int index,
                             int childIdx,
                             Sequence sequence,
                             int sequencePosition)

Description copied from interface: Transition

This method returns the logarithm of the score for the transition.

Specified by:

getLogScoreFor in interface Transition

Parameters:

layer - the layer in the matrix

index - the index encoding the context

childIdx - the index of the child that will be visited

sequence - the sequence

sequencePosition - the position within the sequence

Returns:

the logarithm of the score for the transition

getNumberOfIndexes

public int getNumberOfIndexes(int layer)

Description copied from interface: Transition

This method computes the number of different indexes for a given layer of the matrix.

Specified by:

getNumberOfIndexes in interface Transition

Parameters:

layer - the layer of the matrix

Returns:

the number of different indexes for a given position

hasAnySelfTransitions

public boolean hasAnySelfTransitions()

Description copied from interface: Transition

This method answers the question whether the instance models any self transitions. It returns true if the current instance has any self transitions, otherwise false

Specified by:

hasAnySelfTransitions in interface Transition

Returns:

true if the current instance has any self transitions, otherwise false

getMaximalInDegree

public int getMaximalInDegree()

Description copied from interface: Transition

This method returns the maximal out degree of any context used in this transition instance.

Specified by:

getMaximalInDegree in interface Transition

Returns:

maximal out degree of any context

getMaximalNumberOfChildren

public int getMaximalNumberOfChildren()

Description copied from interface: Transition

This method returns the maximal number of children for any context used in this transition instance.

Specified by:

getMaximalNumberOfChildren in interface Transition

Returns:

maximal number of children for any context

getLastContextState

public int getLastContextState(int layer,
                               int index)

Description copied from interface: Transition

The method returns the index of the state of the context, if there is no context -1 is returned.

Specified by:

getLastContextState in interface Transition

Parameters:

layer - the layer in the matrix

index - the index encoding the context

Returns:

the index of the last state of the context, if there is no context -1 is returned

getChildIdx

public int getChildIdx(int layer,
                       int index,
                       int state)

Description copied from interface: Transition

This method returns the child index of the state, if this state is no child of the context -1 is returned

Specified by:

getChildIdx in interface Transition

Parameters:

layer - the layer in the matrix

index - the index encoding the context

state - the index of the state

Returns:

the child index of the state, if this state is no child of the context -1 is returned

getLogGammaScoreFromStatistic

public double getLogGammaScoreFromStatistic()

Description copied from interface: TransitionWithSufficientStatistic

This method calculates a score for the current statistics, which is independent from the current parameters In general the gamma-score is a product of gamma-functions parameterized with the current statistics

Specified by:

getLogGammaScoreFromStatistic in interface TransitionWithSufficientStatistic

Returns:

the logarithm of the gamma-score for the current statistics

toString

public String toString()

Overrides:

toString in class Object

toString

public String toString(String[] stateNames)

Description copied from interface: Transition

This method returns a String representation of the Transition using the given names of the states.

Specified by:

toString in interface Transition

Parameters:

stateNames - the names of the states, can be null

Returns:

a String representation of the Transition using the given names of the states

isAbsoring

public boolean[] isAbsoring()

Description copied from interface: Transition

This method returns for each state whether it is absorbing or not.

Specified by:

isAbsoring in interface Transition

Returns:

for each state whether it is absorbing or not

setParameters

public void setParameters(Transition t)
                   throws IllegalArgumentException

Description copied from interface: Transition

Set values of parameters of the instance to the value of the parameters of the given instance. It can be assumed that the given instance and the current instance are from the same class. This method might be used for instance in a multi-threaded optimization to broadcast the parameters.

Specified by:

setParameters in interface Transition

Parameters:

t - the transition with the parameters to be set

Throws:

IllegalArgumentException - if the assumption about the same class for given and current instance is wrong