84 lines
2.6 KiB
Python
84 lines
2.6 KiB
Python
|
import numpy as np
|
||
|
|
from numpy.testing import assert_allclose
|
||
|
|
import pytest
|
||
|
|
from scipy.special._ufuncs import _cosine_cdf, _cosine_invcdf
|
||
|
|
|
||
|
|
|
||
|
|
# These values are (x, p) where p is the expected exact value of
|
||
|
|
# _cosine_cdf(x). These values will be tested for exact agreement.
|
||
|
|
_coscdf_exact = [
|
||
|
|
(-4.0, 0.0),
|
||
|
|
(0, 0.5),
|
||
|
|
(np.pi, 1.0),
|
||
|
|
(4.0, 1.0),
|
||
|
|
]
|
||
|
|
|
||
|
|
@pytest.mark.parametrize("x, expected", _coscdf_exact)
|
||
|
|
def test_cosine_cdf_exact(x, expected):
|
||
|
|
assert _cosine_cdf(x) == expected
|
||
|
|
|
||
|
|
|
||
|
|
# These values are (x, p), where p is the expected value of
|
||
|
|
# _cosine_cdf(x). The expected values were computed with mpmath using
|
||
|
|
# 50 digits of precision. These values will be tested for agreement
|
||
|
|
# with the computed values using a very small relative tolerance.
|
||
|
|
# The value at -np.pi is not 0, because -np.pi does not equal -π.
|
||
|
|
_coscdf_close = [
|
||
|
|
(3.1409, 0.999999999991185),
|
||
|
|
(2.25, 0.9819328173287907),
|
||
|
|
# -1.6 is the threshold below which the Pade approximant is used.
|
||
|
|
(-1.599, 0.08641959838382553),
|
||
|
|
(-1.601, 0.086110582992713),
|
||
|
|
(-2.0, 0.0369709335961611),
|
||
|
|
(-3.0, 7.522387241801384e-05),
|
||
|
|
(-3.1415, 2.109869685443648e-14),
|
||
|
|
(-3.14159, 4.956444476505336e-19),
|
||
|
|
(-np.pi, 4.871934450264861e-50),
|
||
|
|
]
|
||
|
|
|
||
|
|
@pytest.mark.parametrize("x, expected", _coscdf_close)
|
||
|
|
def test_cosine_cdf(x, expected):
|
||
|
|
assert_allclose(_cosine_cdf(x), expected, rtol=5e-15)
|
||
|
|
|
||
|
|
|
||
|
|
# These values are (p, x) where x is the expected exact value of
|
||
|
|
# _cosine_invcdf(p). These values will be tested for exact agreement.
|
||
|
|
_cosinvcdf_exact = [
|
||
|
|
(0.0, -np.pi),
|
||
|
|
(0.5, 0.0),
|
||
|
|
(1.0, np.pi),
|
||
|
|
]
|
||
|
|
|
||
|
|
@pytest.mark.parametrize("p, expected", _cosinvcdf_exact)
|
||
|
|
def test_cosine_invcdf_exact(p, expected):
|
||
|
|
assert _cosine_invcdf(p) == expected
|
||
|
|
|
||
|
|
|
||
|
|
def test_cosine_invcdf_invalid_p():
|
||
|
|
# Check that p values outside of [0, 1] return nan.
|
||
|
|
assert np.isnan(_cosine_invcdf([-0.1, 1.1])).all()
|
||
|
|
|
||
|
|
|
||
|
|
# These values are (p, x), where x is the expected value of _cosine_invcdf(p).
|
||
|
|
# The expected values were computed with mpmath using 50 digits of precision.
|
||
|
|
_cosinvcdf_close = [
|
||
|
|
(1e-50, -np.pi),
|
||
|
|
(1e-14, -3.1415204137058454),
|
||
|
|
(1e-08, -3.1343686589124524),
|
||
|
|
(0.0018001, -2.732563923138336),
|
||
|
|
(0.010, -2.41276589008678),
|
||
|
|
(0.060, -1.7881244975330157),
|
||
|
|
(0.125, -1.3752523669869274),
|
||
|
|
(0.250, -0.831711193579736),
|
||
|
|
(0.400, -0.3167954512395289),
|
||
|
|
(0.419, -0.25586025626919906),
|
||
|
|
(0.421, -0.24947570750445663),
|
||
|
|
(0.750, 0.831711193579736),
|
||
|
|
(0.940, 1.7881244975330153),
|
||
|
|
(0.9999999996, 3.1391220839917167),
|
||
|
|
]
|
||
|
|
|
||
|
|
@pytest.mark.parametrize("p, expected", _cosinvcdf_close)
|
||
|
|
def test_cosine_invcdf(p, expected):
|
||
|
|
assert_allclose(_cosine_invcdf(p), expected, rtol=1e-14)
|