def train_hmm_and_keep_track_of_log_likelihood(hmm, obs, n_iter=1, **kwargs): hmm.fit(obs, n_iter=1, **kwargs) loglikelihoods = [] for n in xrange(n_iter): hmm.fit(obs, n_iter=1, init_params='', **kwargs) loglikelihoods.append(sum(hmm.score(x) for x in obs)) return loglikelihoods
def test_hmms(hmms, chain, trajectory): """Returns the scores for a set of hmms over a single trajectory. Parameters: hmms: a dictionary where the keys are the names for the hmms the the values are MultinomialHMM instances. chain: the pre-processing chain created with create_preprocessing_chain.py trajectory: a [n_points, n_features] matrix to be preprocessed by chain. """ input_traj = chain.transform(trajectory) scores = {} for name, hmm in hmms.iteritems(): scores[name] = hmm.score([input_traj]) return scores
def test_hmms(hmms, pca, kmeans, trajectory): input_traj = kmeans.predict(pca.transform(trajectory)) scores = {} for name, hmm in hmms.iteritems(): scores[name] = hmm.score( [input_traj]) return scores
def test_hmms(hmms, pca, kmeans, trajectory): input_traj = kmeans.predict(pca.transform(trajectory)) scores = {} for name, hmm in hmms.iteritems(): scores[name] = hmm.score([input_traj]) return scores