def test_log_marginal_likelyhood_exact():
def exact_function(counts, alpha, beta):
counts_facproduct = functools.reduce(operator.mul,
map(np.math.factorial, counts), 1)
cs = sum(counts)
ns = len(counts)
return np.log((beta**alpha) * math.gamma(cs + alpha) /
(math.gamma(alpha) * counts_facproduct *
((ns + beta)**(cs + alpha))))
scorer = pasio.LogMarginalLikelyhoodComputer(np.array([0]), 3, 5, None)
assert np.allclose(scorer.log_marginal_likelyhood(),
exact_function(np.array([0]), 3, 5))
scorer = pasio.LogMarginalLikelyhoodComputer(np.array([0, 1]), 3, 5, None)
assert np.allclose(scorer.log_marginal_likelyhood(),
exact_function(np.array([0, 1]), 3, 5))
scorer = pasio.LogMarginalLikelyhoodComputer(np.array([4, 0, 1, 3]), 5, 2,
None)
assert np.allclose(scorer.log_marginal_likelyhood(),
exact_function(np.array([4, 0, 1, 3]), 5, 2))
scorer = pasio.LogMarginalLikelyhoodComputer(np.array([4, 0, 1, 3]), 1, 1,
None)
assert np.allclose(scorer.log_marginal_likelyhood(),
exact_function(np.array([4, 0, 1, 3]), 1, 1))
def test_stat_split_into_segments_square():
def split_on_two_segments_or_not(counts, scorer_factory):
scorer = scorer_factory(counts)
best_score = scorer.score(0, len(counts))
split_point = 0
for i in range(len(counts)):
current_score = scorer.score(stop=i) + scorer.score(start=i)
if current_score > best_score:
split_point = i
best_score = current_score
return best_score, split_point
np.random.seed(4)
scorer_factory = lambda counts, split_candidates=None: pasio.LogMarginalLikelyhoodComputer(
counts, 1, 1, split_candidates)
for repeat in range(5):
counts = np.concatenate(
[np.random.poisson(15, 100),
np.random.poisson(20, 100)])
optimal_split = pasio.SquareSplitter().split(counts, scorer_factory)
two_split = split_on_two_segments_or_not(counts, scorer_factory)
assert optimal_split[0] >= two_split[0]
assert two_split[1] in optimal_split[1]
assert np.allclose(
optimal_split[0],
pasio.compute_score_from_splits(counts, optimal_split[1],
scorer_factory))
if (two_split[1] is None):
assert optimal_split[1] == [0, 200]
else:
assert abs(two_split[1] - 100) < 10
def test_benchmark_log_marginal_likehood(benchmark):
counts = np.concatenate(
[np.random.poisson(200, 50),
np.random.poisson(20, 50)])
scorer = pasio.LogMarginalLikelyhoodComputer(counts, 1, 1)
result = benchmark(compute_log_marginal_likelyhood2, scorer, len(counts))
def test_benchmark_segmentation_long(benchmark):
np.random.seed(2)
counts = np.concatenate(
[np.random.poisson(15, 500),
np.random.poisson(20, 500)])
scorer_factory = lambda counts, split_candidates=None: pasio.LogMarginalLikelyhoodComputer(
counts, 1, 1, split_candidates)
result = benchmark(segmentation, counts, scorer_factory)
def test_suffixes_scores():
np.random.seed(2)
counts = np.concatenate(
[np.random.poisson(15, 100),
np.random.poisson(20, 100)])
scorer = pasio.LogMarginalLikelyhoodComputer(counts, 1, 1)
suffixes_scores = [scorer.self_score(i, 150) for i in range(150)]
assert np.allclose(scorer.all_suffixes_self_score(150),
np.array(suffixes_scores))
counts = np.array([0, 0, 1, 0, 0, 2, 2, 2, 10, 11, 100, 1, 0, 0, 1, 0],
dtype='int64')
scorer = pasio.LogMarginalLikelyhoodComputer(counts, 1, 1)
suffixes_scores = [
scorer.self_score(i,
len(counts) - 1) for i in range(len(counts) - 1)
]
assert np.allclose(scorer.all_suffixes_self_score(len(counts) - 1),
np.array(suffixes_scores))
def test_suffixes_scores_with_candidates():
np.random.seed(2)
counts = np.arange(1, 10)
scorer = pasio.LogMarginalLikelyhoodComputer(counts, 1, 1)
candidates = np.array([0, 1, 3, 4, 5, 6, 7, 8, 9])
scorer_with_candidates = pasio.LogMarginalLikelyhoodComputer(
counts, 1, 1, split_candidates=candidates)
candidate_suffixes = scorer.all_suffixes_self_score(9)[candidates[:-1]]
suffixes_just_candidates = scorer_with_candidates.all_suffixes_self_score(
8)
assert np.allclose(candidate_suffixes, suffixes_just_candidates)
counts = np.concatenate(
[np.random.poisson(15, 100),
np.random.poisson(20, 100)])
scorer = pasio.LogMarginalLikelyhoodComputer(counts, 1, 1)
candidates = np.array([0, 1, 10, 20, 21, 30, 40, 149])
scorer_with_candidates = pasio.LogMarginalLikelyhoodComputer(
counts, 1, 1, split_candidates=candidates)
candidate_suffixes = scorer.all_suffixes_self_score(149)[candidates[:-1]]
suffixes_just_candidates = scorer_with_candidates.all_suffixes_self_score(
len(candidates) - 1)
assert np.allclose(candidate_suffixes, suffixes_just_candidates)