Skip to content

Add ICDF and CDF to BernoulliProbs #1652

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

Add ICDF and CDF to BernoulliProbs #1652

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
abb2175
Add ICDF and CDF to BernoulliProbs
12kleingordon34 Sep 27, 2023
3ce5ea4
Add tests for BernoulliProbs ICDF and CDF
12kleingordon34 Sep 27, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions numpyro/distributions/discrete.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,6 +102,13 @@ def mean(self):
def variance(self):
return self.probs * (1 - self.probs)

def cdf(self, value):
return ((1 - self.probs) * jnp.heaviside(value, 1)
+ self.probs * jnp.heaviside(value - 1, 1))

def icdf(self, q):
return jnp.heaviside(q - (1 - self.probs), 1)

def enumerate_support(self, expand=True):
values = jnp.arange(2).reshape((-1,) + (1,) * len(self.batch_shape))
if expand:
Expand Down
13 changes: 10 additions & 3 deletions test/test_distributions.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1387,7 +1387,7 @@ def test_mixture_log_prob():
@pytest.mark.parametrize(
"jax_dist, sp_dist, params",
# TODO: add more complete pattern for Discrete.cdf
CONTINUOUS + [T(dist.Poisson, 2.0), T(dist.Poisson, np.array([2.0, 3.0, 5.0]))],
CONTINUOUS + [T(dist.Poisson, 2.0), T(dist.Poisson, np.array([2.0, 3.0, 5.0])), T(dist.BernoulliProbs, 0.2)],
)
@pytest.mark.filterwarnings("ignore:overflow encountered:RuntimeWarning")
def test_cdf_and_icdf(jax_dist, sp_dist, params):
Expand All @@ -1411,8 +1411,15 @@ def test_cdf_and_icdf(jax_dist, sp_dist, params):
atol=1e-5,
rtol=rtol,
)
assert_allclose(d.cdf(d.icdf(quantiles)), quantiles, atol=1e-5, rtol=1e-5)
assert_allclose(d.icdf(d.cdf(samples)), samples, atol=1e-5, rtol=rtol)
if jax_dist is dist.BernoulliProbs:
assert pytest.approx(d.icdf(quantiles).mean(), abs=0.1) == d.probs
prop_of_ones = (d.cdf(d.icdf(quantiles)) == 1).mean()
prop_of_zeros = (d.cdf(d.icdf(quantiles)) == (1 - d.probs)).mean()
assert pytest.approx(prop_of_ones, abs=0.1) == d.probs
assert pytest.approx(prop_of_zeros, abs=0.1) == (1 - d.probs)
else:
assert_allclose(d.cdf(d.icdf(quantiles)), quantiles, atol=1e-5, rtol=1e-5)
assert_allclose(d.icdf(d.cdf(samples)), samples, atol=1e-5, rtol=rtol)
except NotImplementedError:
pass

Expand Down