def train_N_state_hmms_from_data(filename, num_states, debug=False):
""" reads all the data, then split it up into each category, and then
builds a separate hmm for each category in data """
dataset = DataSet(filename)
category_seqs = split_into_categories(dataset)
# Build a hmm for each category in data
hmms = {}
for cat, seqs in category_seqs.items():
if debug:
print "\n\nLearning %s-state HMM for category %s" % (
num_states, cat)
model = HMM(range(num_states), dataset.outputs)
model.learn_from_observations(seqs, debug)
hmms[cat] = model
if debug:
print "The learned model for %s:" % cat
print model
return (hmms, dataset)
def task(self):
num_states = range(1, MAX_NUM_HIDDEN_STATES)
filename = "weather_bos_la.data"
dataset = DataSet(filename)
category_seqs = split_into_categories(dataset)
boston_seqs = category_seqs["boston"]
likelihoods = []
for N in num_states:
model = HMM(range(N), dataset.outputs)
ll = model.learn_from_observations(boston_seqs, False, True)
likelihoods.append(ll[-1])
chart = {"chart": {"defaultSeriesType": "line"},
"xAxis": {"title": {"text": "number of hidden states"},
"categories": num_states},
"yAxis": {"title": {"text": "Fraction Correct"}},
"title": {"text": "log likelihood of HMMs"
" modeling boston weather"},
"series": [{"name": "boston training data",
"data": likelihoods}]}
return chart
