def _initializeGoodModels(data, numModels, states):
"""Initialization technique for selecting a good pool of models by which
to start the expectation maximization algorithm for unsupervised learning
Based on kmeans++ paper.
"""
models = []
distances = numpy.zeros((len(data)), float)
#Select initial random sequence
seq = data[int(random.random() * len(data))]
models.append(hmmsup.obsToModel(seq))
for i in range(numModels - 1):
for j in range(len(data)):
tmp = hmmsup.obsToModel(data[j])
tmpDistances = hmmsup.hmmDistAll(data[j], models)
tmpDistances.sort()
distances[j] = tmpDistances[0]
distances /= sum(distances)
#Select a model with probability dist/sum(all distance)
val = random.random()
distSum = 0
for j in range(len(distances)):
distSum += distances[j]
if distSum >= val:
#Make a new model using this data element
models.append(hmmsup.obsToModel(data[j]))
break
return models
def _initializeGoodModels(data, numModels, states):
"""Initialization technique for selecting a good pool of models by which
to start the expectation maximization algorithm for unsupervised learning
Based on kmeans++ paper.
"""
models = []
distances = numpy.zeros((len(data)), float)
#Select initial random sequence
seq = data[int(random.random()*len(data))]
models.append(hmmsup.obsToModel(seq))
for i in range(numModels - 1):
for j in range(len(data)):
tmp = hmmsup.obsToModel(data[j])
tmpDistances = hmmsup.hmmDistAll(data[j], models)
tmpDistances.sort()
distances[j] = tmpDistances[0]
distances /= sum(distances)
#Select a model with probability dist/sum(all distance)
val = random.random()
distSum = 0
for j in range(len(distances)):
distSum += distances[j]
if distSum >= val:
#Make a new model using this data element
models.append(hmmsup.obsToModel(data[j]))
break
return models
def _trainModels(tdata, models):
"""Train models using every data element designated from the _assign
functions.
Note: this function is independent from the type of data split used.
"""
for i in range(len(models)):
#Create a sequence set used for training from the multiple observations
seqSet = ghmm.SequenceSet(ghmm.Float(), [])
for tmpData in tdata[i]:
seqSet.merge(ghmm.EmissionSequence(ghmm.Float(), tmpData))
#Make average sequence
s = numpy.array(tdata[i])
nm = hmmsup.obsToModel(s.mean(axis=0), max(s.std(axis=0)))
nm.normalize()
nm.baumWelch(seqSet)
models[i] = nm
#models[i].baumWelch(seqSet)#, loglikelihoodCutoff = 0.000001)
hmmsup.normalizeAMat(models[i])
hmmsup.normalizePiMat(models[i])
return models
def _trainModels(tdata, models):
"""Train models using every data element designated from the _assign
functions.
Note: this function is independent from the type of data split used.
"""
for i in range(len(models)):
#Create a sequence set used for training from the multiple observations
seqSet = ghmm.SequenceSet(ghmm.Float(), [])
for tmpData in tdata[i]:
seqSet.merge(ghmm.EmissionSequence(ghmm.Float(), tmpData))
#Make average sequence
s = numpy.array(tdata[i])
nm = hmmsup.obsToModel(s.mean(axis = 0), max(s.std(axis = 0)))
nm.normalize()
nm.baumWelch(seqSet)
models[i] = nm
#models[i].baumWelch(seqSet)#, loglikelihoodCutoff = 0.000001)
hmmsup.normalizeAMat(models[i])
hmmsup.normalizePiMat(models[i])
return models
