def test_hk_big_sigma(pts, dims):
"""We expect that with a huge sigma, the diagrams are so diluted that
they are almost 0. Effectively, verifies that the smoothing is applied."""
n_bins = 10
x = get_input(pts, dims)
hk = HeatKernel(sigma=100 * np.max(np.abs(x)), n_bins=n_bins)
x_t = hk.fit_transform(x)
assert np.all(np.abs(x_t) <= 1e-4)
def test_hk_shape(n_jobs, pts, dims):
n_bins = 10
X = get_input(pts, dims)
sigma = (np.max(X[:, :, :2]) - np.min(X[:, :, :2])) / 2
hk = HeatKernel(sigma=sigma, n_bins=n_bins, n_jobs=n_jobs)
num_dimensions = len(np.unique(dims))
X_t = hk.fit_transform(X)
assert X_t.shape == (X.shape[0], num_dimensions, n_bins, n_bins)
def test_hk_positive(pts, dims):
"""We expect the points above the PD-diagonal to be non-negative (up to a
numerical error)"""
n_bins = 10
X = get_input(pts, dims)
sigma = (np.max(X[:, :, :2]) - np.min(X[:, :, :2])) / 2
hk = HeatKernel(sigma=sigma, n_bins=n_bins)
X_t = hk.fit_transform(X)
assert np.all((np.tril(X_t[:, :, ::-1, :]) + 1e-13) >= 0.)
def test_large_hk_shape_parallel():
"""Test that HeatKernel returns something of the right shape when the input
array is at least 1MB and more than 1 process is used, triggering joblib's
use of memmaps"""
X = np.linspace(0, 100, 300000)
n_bins = 10
diagrams = np.expand_dims(np.stack([X, X, np.zeros(len(X))]).transpose(),
axis=0)
hk = HeatKernel(sigma=1, n_bins=n_bins, n_jobs=2)
num_dimensions = 1
x_t = hk.fit_transform(diagrams)
assert x_t.shape == (diagrams.shape[0], num_dimensions, n_bins, n_bins)
def get_heat_kernel(persistence_diagram):
pi = HeatKernel(sigma=0.001, n_bins=N_BINS, n_jobs=N_JOBS)
print("Computed heat kernel")
return pi.fit_transform(persistence_diagram)
