def main(argv=None):
if argv is None:
argv = sys.argv
usage = "usage: %prog [options] N datafile (pass -h for more info)"
parser = OptionParser(usage)
parser.add_option("-v", "--verbose",
action="store_true", dest="verbose", default=False,
help="Print extra debugging info")
(options, args) = parser.parse_args(argv[1:])
if len(args) != 2:
print "ERROR: Missing arguments"
parser.print_usage()
sys.exit(1)
num_states = int(args[0])
filename = args[1]
filename = normalize_filename(filename)
# Read all the data, then split it up into each category
# Build models from the category data files
hmms, dataset = train_N_state_hmms_from_data(
filename, num_states, options.verbose)
# See how well we do in classifying test sequences
fraction_incorrect = compute_classification_performance(
hmms, dataset, options.verbose)
print(fraction_incorrect)
return 0
def rename_by_taginfo(self, scheme=default_scheme):
taginfo = self.extract_taginfo()
filename = "%s%s" % ( eval_scheme(scheme, taginfo),
self.extension,
)
filename_good = normalize_filename(filename)
infomsg ("filename_good: %s => %s" % (self.filename, filename_good))
os.rename(self.filename, filename_good)
self.filename = filename_good
self.basename = os.path.splitext(filename_good)[0]
def train_hmm_from_data(data_filename, debug=False):
if debug:
print "\n\nReading dataset %s ..." % data_filename
data_filename = normalize_filename(data_filename)
d = DataSet(data_filename)
# if options.verbose:
# print d
if debug:
print "Building an HMM from the full training data..."
hmm = HMM(d.states, d.outputs)
hmm.learn_from_labeled_data(d.train_state, d.train_output)
if debug:
print "The model:"
print hmm
return (hmm, d)
def test_small_robot_dataset(self):
data_filename = "robot_small.data"
data_filename = normalize_filename(data_filename)
hmm, d = train_hmm_from_data(data_filename)
err_full = run_viterbi(hmm, d, True)
self.assertAlmostEqual(err_full, 2.0 / 9)
def test_normalize_filename(filename: str, result: str):
assert util.normalize_filename(filename) == result
