def test_jsd(self):
"""evaluate jsd between identical, and non-identical distributions"""
# case1 is testing if the jsd between two identical distributions is 0.0
case1 = [
[0.0, 0.0, 0.0],
[0.0, 0.0, 0.0],
]
for index in range(len(case1[0])):
case1[0][index] = 1.0
case1[1][index] = 1.0
assert_allclose(
jsd(case1[0], case1[1], validate=True),
0.0,
err_msg="Testing case1 for jsd failed",
atol=self.atol,
)
case1[0][index] = 0.0
case1[1][index] = 0.0
# case2 is testing the numerical output of jsd between two distant distributions
case2 = [[1 / 10, 9 / 10, 0], [0, 1 / 10, 9 / 10]]
assert_allclose(
jsd(case2[0], case2[1], validate=True),
0.7655022032053593,
err_msg="Testing case2 for jsd failed",
atol=self.atol,
)
# case3 is testing the numerical output of jsd between two distant distributions
case3 = [[1.0, 0.0], [1 / 2, 1 / 2]]
assert_allclose(
jsd(case3[0], case3[1], validate=True),
0.3112781244591328,
err_msg="Testing case3 for jsd failed",
atol=self.atol,
)
# case4 - the jsd between two identical uniform distributions is 0.0
case4 = [
[1 / 10] * 10,
[1 / 10] * 10,
]
assert_allclose(
jsd(case4[0], case4[1], validate=True),
0.0,
err_msg="Testing case4 for jsd failed",
atol=self.atol,
)
assert_allclose(
jsd(case4[0], case4[0], validate=True),
0.0,
err_msg="Testing case4 for jsd failed",
atol=self.atol,
)
def test_jsd(self):
"""case1 is testing if the jsd between two identical distributions is 0.0"""
case1 = [
[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
]
for pointer in range(10):
case1[0][pointer] = 1.0
case1[1][pointer] = 1.0
assert_allclose(
jsd(case1[0], case1[1], validate=True),
0.0,
err_msg="Testing case1 for jsd failed",
)
case1[0][pointer] = 0.0
case1[1][pointer] = 0.0
"""case2 is testing the numerical output of jsd between two random distributions"""
case2 = [[1.0 / 10, 9.0 / 10, 0], [0, 1.0 / 10, 9.0 / 10]]
assert_allclose(
jsd(case2[0], case2[1], validate=True),
0.7655022032053593,
err_msg="Testing case2 for jsd failed",
)
"""case3 is testing the numerical output of jsd between two random distributions"""
case3 = [[1.0, 0.0], [0.5, 0.5]]
assert_allclose(
jsd(case3[0], case3[1], validate=True),
0.3112781244591328,
err_msg="Testing case3 for jsd failed",
)
"""case4 is testing if the jsd between two identical uniform distributions is 0.0"""
case4 = [
[0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
[0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1],
]
assert_allclose(
jsd(case4[0], case4[1], validate=True),
0.0,
err_msg="Testing case4 for jsd failed",
)
def test_jsd_precision(self):
"""handle case where the difference is incredibly small"""
pi_0 = [
0.4398948756903677,
0.1623791467423164,
0.31844113569205656,
0.07928484187525932,
]
pi_1 = [
0.43989487569036767,
0.16237914674231643,
0.3184411356920566,
0.07928484187525933,
]
result = jsd(pi_0, pi_1)
self.assertTrue(result >= 0)
def test_pairwise_jsd(self):
"""correctly constructs pairwise JSD dict"""
from numpy.random import random
from cogent3.maths.measure import jsd
data = [[0.25, 0.25, 0.25, 0.25], [0.5, 0.5, 0, 0]]
expect = jsd(data[0], data[1])
freqs = MotifFreqsArray(array(data), "ACGT")
got = freqs.pairwise_jsd()
assert_allclose(list(got.values())[0], expect)
data = []
for _ in range(6):
freqs = random(4)
freqs = freqs / freqs.sum()
data.append(freqs)
freqs = MotifFreqsArray(array(data), "ACGT")
pwise = freqs.pairwise_jsd()
self.assertEqual(len(pwise), 6 * 6 - 6)
def test_jsd_validation(self):
"""jsd fails with malformed data"""
freqs1 = random(5)
normalised_freqs1 = freqs1 / freqs1.sum()
two_dimensional_freqs1 = [freqs1, freqs1]
shorter_freqs1 = freqs1[:4]
freqs2 = random(5)
normalised_freqs2 = freqs2 / freqs2.sum()
two_dimensional_freqs2 = [freqs2, freqs2]
shorter_freqs2 = freqs2[:4]
with self.assertRaises(AssertionError):
jsd(
freqs1, two_dimensional_freqs2, validate=True
) # freqs1/freqs2 mismatched shape
with self.assertRaises(AssertionError):
jsd(
two_dimensional_freqs1, freqs2, validate=True
) # freqs1/freqs2 mismatched shape
with self.assertRaises(AssertionError):
jsd(freqs1, shorter_freqs2, validate=True) # freqs1/freqs2 mismatched shape
with self.assertRaises(AssertionError):
jsd(shorter_freqs1, freqs2, validate=True) # freqs1/freqs2 mismatched shape
with self.assertRaises(AssertionError):
jsd(
two_dimensional_freqs1, freqs2, validate=True
) # freqs1 has incorrect dimension
with self.assertRaises(AssertionError):
jsd(
two_dimensional_freqs1, two_dimensional_freqs2, validate=True
) # freqs1 has incorrect dimension
with self.assertRaises(AssertionError):
jsd(
freqs1, two_dimensional_freqs2, validate=True
) # freqs2 has incorrect dimension
with self.assertRaises(AssertionError):
jsd(freqs1, freqs2, validate=True) # invalid freqs1
with self.assertRaises(AssertionError):
jsd(freqs1, normalised_freqs2, validate=True) # invalid freqs1
with self.assertRaises(AssertionError):
jsd(normalised_freqs1, freqs2, validate=True) # invalid freqs2