def make_data_banded(n_samples, n_features):
alpha = 0.1
cov, prec, adj = LatticeGraph(n_blocks=2,
random_sign=True,
chain_blocks=True,
seed=1).create(n_features, alpha)
prng = np.random.RandomState(2)
X = prng.multivariate_normal(np.zeros(n_features), cov, size=n_samples)
return X, cov, prec
class TestGraphs(object):
def test_lattice_random_sign_false(self):
n_features = 10
alpha = 0.2
prng = np.random.RandomState(1)
adjacency = lattice(prng, n_features, alpha, random_sign=False)
nnz = alpha * n_features
# diagonal zero
assert np.sum(adjacency[np.where(np.eye(n_features))]) == 0
# all non-zeros are negative (random_sign=False)
assert np.sum(adjacency[adjacency != 0] < 0) == np.sum(
adjacency[adjacency != 0] != 0)
# banded structure
for nn in range(n_features):
d = np.diag(np.ones((n_features - nn, )), k=nn)
if nn == 0:
continue
elif nn <= nnz:
assert np.sum(adjacency[np.where(d)]) != 0
else:
assert np.sum(adjacency[np.where(d)]) == 0
def test_lattice_random_sign_true(self):
n_features = 10
alpha = 0.2
prng = np.random.RandomState(1)
adjacency = lattice(prng, n_features, alpha, random_sign=True)
nnz = alpha * n_features
# diagonal zero
assert np.sum(adjacency[np.where(np.eye(n_features))]) == 0
# some non-zeros are negative (random_sign=True)
assert np.sum(adjacency[adjacency != 0] < 0) != np.sum(
adjacency[adjacency != 0] != 0)
# banded structure
for nn in range(n_features):
d = np.diag(np.ones((n_features - nn, )), k=nn)
if nn == 0:
continue
elif nn <= nnz:
assert np.sum(adjacency[np.where(d)]) != 0
else:
assert np.sum(adjacency[np.where(d)]) == 0
def test_blocks_chain_blocks_false(self):
prng = np.random.RandomState(1)
n_block_features = 4
n_blocks = 3
block = np.ones((n_block_features, n_block_features))
adjacency = blocks(prng, block, n_blocks=n_blocks, chain_blocks=False)
n_features = n_block_features * n_blocks
assert adjacency.shape[0] == n_features
# diagonal zero
assert np.sum(adjacency[np.where(np.eye(n_features))]) == 0
# for each block, assert sum of the blocks minus the diagonal
for nn in range(n_blocks):
brange = slice(nn * n_block_features, (nn + 1) * n_block_features)
assert (np.sum(adjacency[brange, brange]) == n_block_features**2 -
n_block_features)
# assert that all nonzeros equal sum of all blocks above
assert (np.sum(
adjacency.flat) == (n_block_features**2 - n_block_features) *
n_blocks)
def test_blocks_chain_blocks_true(self):
prng = np.random.RandomState(1)
n_block_features = 4
n_blocks = 3
block = np.ones((n_block_features, n_block_features))
adjacency = blocks(prng, block, n_blocks=n_blocks, chain_blocks=True)
n_features = n_block_features * n_blocks
assert adjacency.shape[0] == n_features
# diagonal zero
assert np.sum(adjacency[np.where(np.eye(n_features))]) == 0
# for each block, assert sum of the blocks minus the diagonal
for nn in range(n_blocks):
brange = slice(nn * n_block_features, (nn + 1) * n_block_features)
assert (np.sum(adjacency[brange, brange]) == n_block_features**2 -
n_block_features)
# assert that all nonzeros DO NOT equal sum of all blocks above
assert (np.sum(adjacency.flat) !=
(n_block_features**2 - n_block_features) * n_blocks)
@pytest.mark.parametrize(
"graph,seed",
[
(ClusterGraph(), None),
(ErdosRenyiGraph(), None),
(LatticeGraph(), None),
(LatticeGraph(random_sign=True, chain_blocks=False, seed=3), 3),
(ClusterGraph(n_blocks=5, seed=3), 3),
(ErdosRenyiGraph(seed=3), 3),
],
)
def test_graph_classes(self, graph, seed):
"""Simple smell test on inidivudal graph classes"""
n_features = 50
alpha = 0.1
covariance, precision, adjacency = graph.create(n_features, alpha)
assert covariance.shape[0] == covariance.shape[1]
assert covariance.shape[0] == n_features
assert precision.shape[0] == precision.shape[1]
assert precision.shape[0] == n_features
assert adjacency.shape[0] == adjacency.shape[1]
assert adjacency.shape[0] == n_features
assert np.sum(covariance.flat) > 0
assert np.sum(precision.flat) > 0
assert np.sum(adjacency.flat) > 0
if seed is not None:
assert graph.seed == seed
metrics = {
"fp_rate": support_false_positive_count,
"fn_rate": support_false_negative_count,
"support_error": support_difference_count,
"prob_exact_support": has_exact_support,
"prob_approx_support": has_approx_support,
"frobenius": error_fro,
}
###############################################################################
# Run MC trials
mc = MonteCarloProfile(
n_features=50,
n_trials=10,
graph=LatticeGraph(),
n_samples_grid=10,
alpha_grid=5,
metrics=metrics,
verbose=True,
n_jobs=4,
)
mc.fit()
###############################################################################
# Plot results for each metric
for key in metrics:
plt.figure()
plt.plot(mc.grid_, mc.results_[key].T, linewidth=2)
plt.title("metric = {}".format(key))