def test_state_priors_weights_array_pseudocounts(self):
pmat = numpy.random.randint(0, high=255, size=(self.num_states, ))
my_counts = numpy.zeros(self.num_states)
for my_seq, my_weight in zip(self.test_seqs, self.test_weights):
my_counts[my_seq[0]] += my_weight
expected_prior_counts = my_counts + pmat
expected_prior_freqs = (
1.0 * expected_prior_counts) / expected_prior_counts.sum()
found_prior_counts, found_transition_counts = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
state_prior_pseudocounts=pmat,
normalize=False)
found_prior_freqs, found_transition_freqs = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
state_prior_pseudocounts=pmat,
normalize=True)
yield check_tuple_equal, found_prior_counts.shape, (self.num_states, )
yield check_array_equal, found_prior_counts, expected_prior_counts
yield check_array_equal, found_prior_freqs, expected_prior_freqs
def test_state_priors_weights_int_pseudocounts(self):
my_counts = numpy.zeros(self.num_states)
for (my_seq, my_weight) in zip(self.test_seqs, self.test_weights):
my_counts[my_seq[0]] += my_weight
for pcounts in (1, 2, 3):
expected_prior_counts = my_counts + pcounts
expected_prior_freqs = (
1.0 * expected_prior_counts) / expected_prior_counts.sum()
found_prior_counts, found_transition_counts = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
state_prior_pseudocounts=pcounts,
normalize=False)
found_prior_freqs, found_transition_freqs = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
state_prior_pseudocounts=pcounts,
normalize=True)
yield check_tuple_equal, found_prior_counts.shape, (
self.num_states, )
yield check_array_equal, found_prior_counts, expected_prior_counts
yield check_array_equal, found_prior_freqs, expected_prior_freqs
def test_transitions_weights_array_pseudocounts(self):
pmat = numpy.random.randint(0,
high=255,
size=(self.num_states, self.num_states))
expected_transition_counts = 0
for mat, weight in zip(self.mats, self.test_weights):
expected_transition_counts += weight * mat
expected_transition_counts += pmat
expected_transition_freqs = (1.0 * expected_transition_counts.T /
expected_transition_counts.sum(1)).T
found_prior_counts, found_transition_counts = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
transition_pseudocounts=pmat,
normalize=False)
found_prior_freqs, found_transition_freqs = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
transition_pseudocounts=pmat,
normalize=True)
yield check_tuple_equal, found_transition_counts.shape, (
self.num_states, self.num_states)
yield check_array_equal, found_transition_counts, expected_transition_counts
yield check_array_equal, found_transition_freqs, expected_transition_freqs
def test_transitions_weights_int_pseudocounts(self):
my_counts = 0
for mat, weight in zip(self.mats, self.test_weights):
my_counts += weight * mat
for pcounts in (1, 2, 3):
expected_transition_counts = my_counts + pcounts
expected_transition_freqs = (1.0 * expected_transition_counts.T /
expected_transition_counts.sum(1)).T
found_prior_counts, found_transition_counts = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
transition_pseudocounts=pcounts,
normalize=False)
found_prior_freqs, found_transition_freqs = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
transition_pseudocounts=pcounts,
normalize=True)
yield check_tuple_equal, found_transition_counts.shape, (
self.num_states, self.num_states)
yield check_array_equal, found_transition_counts, expected_transition_counts
yield check_array_equal, found_transition_freqs, expected_transition_freqs
def test_transitions_no_weights_no_pseudocounts(self):
expected_transition_counts = sum(self.mats)
expected_transition_freqs = (1.0 * expected_transition_counts.T /
expected_transition_counts.sum(1)).T
found_prior_counts, found_transition_counts = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=None,
transition_pseudocounts=0,
normalize=False)
found_prior_freqs, found_transition_freqs = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=None,
transition_pseudocounts=0,
normalize=True)
yield check_tuple_equal, found_transition_counts.shape, (
self.num_states, self.num_states)
yield check_array_equal, found_transition_counts, expected_transition_counts
yield check_array_equal, found_transition_freqs, expected_transition_freqs
def test_transitions_alternate_initializer(self):
expected_transition_counts = 0
for mat, weight in zip(self.mats, self.test_weights):
expected_transition_counts += weight * mat
found_prior_counts, found_transition_counts = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=self.test_weights,
normalize=False,
initializer=scipy.sparse.dok_matrix)
found_dense = found_transition_counts.todense()
assert_array_equal(found_dense, expected_transition_counts)
def test_state_priors_no_weights_no_pseudocounts(self):
expected_prior_counts = numpy.zeros(self.num_states)
for my_seq in self.test_seqs:
expected_prior_counts[my_seq[0]] += 1
expected_prior_freqs = (
1.0 * expected_prior_counts) / expected_prior_counts.sum()
found_prior_counts, found_transition_counts = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=None,
transition_pseudocounts=0,
normalize=False)
found_prior_freqs, found_transition_freqs = build_hmm_tables(
self.num_states,
self.test_seqs,
weights=None,
transition_pseudocounts=0,
normalize=True)
yield check_tuple_equal, found_prior_counts.shape, (self.num_states, )
yield check_array_equal, found_prior_counts, expected_prior_counts
yield check_array_equal, found_prior_freqs, expected_prior_freqs