def main(args, letter_to_emission, trans, emiss):
if args.ticks and args.pbar:
raise ValueError('it is silly to use both ticks and a progress bar')
if args.pbar:
pbar = progress.Bar(1 + 1 + args.n)
else:
pbar = None
with open(args.fasta) as fin:
raw_observations = fasta_to_raw_observations(fin.readlines())
arr = [letter_to_emission[c] for c in raw_observations]
observations = np.array(arr, dtype=np.int8)
if pbar:
pbar.increment()
model = estimation.FiniteModel(trans, emiss, observations)
summary = Summary(model, pbar, args.ticks)
estimation.baum_welch(
model.update, model.get_expectations, trans, emiss, args.n,
callback=summary.after_baum_welch)
summary_text = summary.finish()
suffix = args.common_suffix
if args.summary:
with open(args.summary + suffix, 'w') as fout:
print >> fout, summary_text
if args.log_likelihoods:
with open(args.log_likelihoods + suffix, 'w') as fout:
print >> fout, '\n'.join('%f' % x for x in summary.log_likelihoods)
if args.posterior:
hmm.pretty_print_posterior(raw_observations,
model.get_posterior(), 60, args.posterior + suffix)
if args.posterior_decoding:
hmm.pretty_print_posterior_decoding(raw_observations,
model.get_posterior(), 60, args.posterior_decoding + suffix)
def parallel_worker(params):
"""
@param params: a bunch of parameter packed into a sequence
"""
# unpack the params
user_params, hmm_params = params
f_in, d_out, niterations = user_params
letter_to_emission, trans, emiss = hmm_params
trans = np.array(trans)
emiss = np.array(emiss)
# define output path names
suffix = '.' + os.path.basename(f_in) + ('.n%d.txt' % niterations)
summary_name = os.path.join(d_out, 'summary' + suffix)
log_likelihoods_name = os.path.join(d_out, 'log.likelihood' + suffix)
posterior_name = os.path.join(d_out, 'soft.decoding' + suffix)
posterior_decoding_name = os.path.join(d_out, 'hard.decoding' + suffix)
# read the fasta file
with open(f_in) as fin:
raw_observations = fasta_to_raw_observations(fin.readlines())
arr = [letter_to_emission[c] for c in raw_observations]
observations = np.array(arr, dtype=np.int8)
# do the baum welch stuff
model = estimation.FiniteModel(trans, emiss, observations)
summary = Summary(model, None, None)
estimation.baum_welch(
model.update, model.get_expectations, trans, emiss, niterations,
callback=summary.after_baum_welch)
summary_text = summary.finish()
# write the output files
with open(summary_name, 'wt') as fout:
print >> fout, summary_text
with open(log_likelihoods_name, 'wt') as fout:
print >> fout, '\n'.join('%f' % x for x in summary.log_likelihoods)
hmm.pretty_print_posterior(raw_observations,
model.get_posterior(), 60, posterior_name)
hmm.pretty_print_posterior_decoding(raw_observations,
model.get_posterior(), 60, posterior_decoding_name)
