def test_find_binning_thresholds_regular_data():
data = np.linspace(0, 10, 1001)
bin_thresholds = _find_binning_thresholds(data, max_bins=10)
assert_allclose(bin_thresholds, [1, 2, 3, 4, 5, 6, 7, 8, 9])
bin_thresholds = _find_binning_thresholds(data, max_bins=5)
assert_allclose(bin_thresholds, [2, 4, 6, 8])
def test_find_binning_thresholds_small_regular_data():
data = np.linspace(0, 10, 11)
bin_thresholds = _find_binning_thresholds(data, max_bins=5)
assert_allclose(bin_thresholds, [2, 4, 6, 8])
bin_thresholds = _find_binning_thresholds(data, max_bins=10)
assert_allclose(bin_thresholds, [1, 2, 3, 4, 5, 6, 7, 8, 9])
bin_thresholds = _find_binning_thresholds(data, max_bins=11)
assert_allclose(bin_thresholds, np.arange(10) + 0.5)
bin_thresholds = _find_binning_thresholds(data, max_bins=255)
assert_allclose(bin_thresholds, np.arange(10) + 0.5)
def test_find_binning_thresholds_low_n_bins():
bin_thresholds = [
_find_binning_thresholds(DATA[:, i], max_bins=128) for i in range(2)
]
for i in range(len(bin_thresholds)):
assert bin_thresholds[i].shape == (127,) # 128 - 1
assert bin_thresholds[i].dtype == DATA.dtype
def test_find_binning_thresholds_random_data():
bin_thresholds = [
_find_binning_thresholds(DATA[:, i], max_bins=255) for i in range(2)
]
for i in range(len(bin_thresholds)):
assert bin_thresholds[i].shape == (254,) # 255 - 1
assert bin_thresholds[i].dtype == DATA.dtype
assert_allclose(
bin_thresholds[0][[64, 128, 192]], np.array([-0.7, 0.0, 0.7]), atol=1e-1
)
assert_allclose(
bin_thresholds[1][[64, 128, 192]], np.array([9.99, 10.00, 10.01]), atol=1e-2
)
def test_map_to_bins(max_bins):
bin_thresholds = [
_find_binning_thresholds(DATA[:, i], max_bins=max_bins) for i in range(2)
]
binned = np.zeros_like(DATA, dtype=X_BINNED_DTYPE, order="F")
last_bin_idx = max_bins
_map_to_bins(DATA, bin_thresholds, last_bin_idx, binned)
assert binned.shape == DATA.shape
assert binned.dtype == np.uint8
assert binned.flags.f_contiguous
min_indices = DATA.argmin(axis=0)
max_indices = DATA.argmax(axis=0)
for feature_idx, min_idx in enumerate(min_indices):
assert binned[min_idx, feature_idx] == 0
for feature_idx, max_idx in enumerate(max_indices):
assert binned[max_idx, feature_idx] == max_bins - 1
