Table of Contents

Class: DiscreteEmissionHMM /amd/bernoulli/1/home/abt_vin/georgi/hmm/0.7/ghmm//ghmmwrapper/ghmm.py

HMMs with discrete emissions. Optional features: - silent states - higher order states - parameter tying in training - background probabilities in training

Base Classes   
HMM
Methods   
__del__
__init__
__str__
applyBackground
assignAllBackgrounds
assignStateBackground
backwardTermination
baumWelch
extendDurations
 getBackgroundAssignments
getSilentFlag
getTieGroups
isSilent
logprob
normalize
pathPosterior
posterior
removeTiegroups
 setBackground
setEmission
setTieGroups
statePosterior
toMatrices
toXML
updateTieGroups
  __del__  
__del__ ( self )
  __init__  
__init__ (
        self,
        emissionDomain,
        distribution,
        cmodel,
        )
  __str__  
__str__ ( self )
  applyBackground  
applyBackground ( self,  backgroundWeight )

Apply the background distribution to the emission probabilities of states which have been assigned one (usually in the editor and coded in the XML). applyBackground computes a convex combination of the emission probability and the background, where the backgroundWeight parameter (within [0,1]) controls the background's contribution for each state.

  assignAllBackgrounds  
assignAllBackgrounds ( self,  stateBackground )

Change all the assignments of background distributions to states.

Exceptions   
TypeError, "list required got " + str(type( stateBackground ) )
  assignStateBackground  
assignStateBackground (
        self,
        state,
        backgroundID,
        )
  backwardTermination  
backwardTermination (
        self,
        emissionSequence,
        pybeta,
        scalingVector,
        )

Result: the backward log probability of emissionSequence

Exceptions   
TypeError, "EmissionSequence required, got " + str( emissionSequence.__class__.__name__ )
  baumWelch  
baumWelch (
        self,
        trainingSequences,
        nrSteps=None,
        loglikelihoodCutoff=None,
        )

Reestimates the model with the sequence in trainingSequences.

Note that training for models including silent states is not yet supported.

nrSteps is the maximal number of BW-steps loglikelihoodCutoff is the least relative improvement in likelihood with respect to the last iteration required to continue.

Exceptions   
TypeError, "EmissionSequence or SequenceSet required, got " + str( trainingSequences.__class__.__name__ )
  extendDurations  
extendDurations ( self,  durationlist )

extend states with durations larger one this done by explicit state copying in C

  getBackgroundAssignments  
getBackgroundAssignments ( self )
  getSilentFlag  
getSilentFlag ( self,  state )
  getTieGroups  
getTieGroups ( self )
  isSilent  
isSilent ( self,  state )

Returns True if state is silent, False otherwise

  logprob  
logprob (
        self,
        emissionSequence,
        stateSequence,
        )

log P[ emissionSequence, stateSequence| m]

Exceptions   
SequenceCannotBeBuild, "first symbol " + str(emissionSequence [ i + 1 ] ) + " not emitted by state " + str(stateSequence [ 0 ] )
SequenceCannotBeBuild, "symbol " + str(emissionSequence [ i + 1 ] ) + " not emitted by state " + str(stateSequence [ i + 1 ] )
TypeError, "EmissionSequence required, got " + str( emissionSequence.__class__.__name__ )
  normalize  
normalize ( self )

Normalize transition probs, emission probs (if applicable)

  pathPosterior  
pathPosterior (
        self,
        sequence,
        path,
        )

Returns the log posterior probability for path having generated sequence.

CAVEAT: statePosterior needs to calculate the complete forward and backward matrices. If you are interested in multiple paths it would be more efficient to use the posterior function directly and not multiple calls to pathPosterior

Exceptions   
RuntimeError, "Models with silent states not yet supported."
  posterior  
posterior ( self,  sequence )

Posterior distribution matrix for sequence.

Exceptions   
RuntimeError, "Models with silent states not yet supported."
  removeTiegroups  
removeTiegroups ( self )
  setBackground  
setBackground (
        self,
        backgroundObject,
        stateBackground,
        )

Configure model to use the background distributions in backgroundObject. stateBackground is a list of indixes (one for each state) refering to distributions in backgroundObject.

Note: values in backgroundObject are deepcopied into model

Exceptions   
TypeError, "BackgroundDistribution required, got " + str( emissionSequences.__class__.__name__ )
TypeError, "list required got " + str(type( stateBackground ) )
  setEmission  
setEmission (
        self,
        i,
        distributionParameters,
        )

Set the emission distribution parameters for a discrete model.

  setTieGroups  
setTieGroups ( self,  tieList )
  statePosterior  
statePosterior (
        self,
        sequence,
        state,
        time,
        )

Return the log posterior probability for being at state at time time in sequence.

CAVEAT: statePosterior needs to calculate the complete forward and backward matrices. If you are interested in multiple states it would be more efficient to use the posterior function directly and not multiple calls to statePosterior

Exceptions   
RuntimeError, "Models with silent states not yet supported."
  toMatrices  
toMatrices ( self )

Return the parameters in matrix form.

  toXML  
toXML (
        self,
        filename,
        backgroundobj=None,
        )
  updateTieGroups  
updateTieGroups ( self )

Table of Contents

This document was automatically generated on Fri Jan 20 14:56:59 2006 by HappyDoc version WORKING