Implement sparse Poisson.log_prob() #1003
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
fehiepsi
reviewed
Apr 12, 2021
yup, both look good to me. Please go with the one you prefer. :) |
1 task
fritzo
commented
Apr 13, 2021
| @pytest.mark.parametrize( | ||
| "jax_dist, sp_dist, params", CONTINUOUS + DISCRETE + DIRECTIONAL | ||
| ) | ||
| def test_jit_log_likelihood(jax_dist, sp_dist, params): |
There was a problem hiding this comment.
Whereas jitting in test_log_prob fixes params and varies the value, jitting in this test fixes the value and varies params. This test is thus closer to usage in an observe statement numpyro.sample(..., obs=data).
fehiepsi
previously approved these changes
Apr 13, 2021
| shape = lax.broadcast_shapes(self.batch_shape, jnp.shape(value)) | ||
| rate = jnp.broadcast_to(self.rate, shape).reshape(-1) | ||
| nonzero = jnp.broadcast_to(value > 0, shape).reshape(-1).nonzero() | ||
| value = jnp.broadcast_to(value, shape).reshape(-1) |
There was a problem hiding this comment.
nit: I guess we can swap the order of those two lines:
value = jnp.broadcast_to(value, shape).reshape(-1)
nonzero = (value > 0).nonzero()
Btw, do we need to call .nonzero() here?
There was a problem hiding this comment.
Done. I think .nonzero() was required for tests to pass in an earlier commit, but now things pass without it, so I've removed.
fehiepsi
approved these changes
Apr 13, 2021
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This adds an
is_sparsekwarg todist.Poissontriggering a cheaper.log_prob()computaiton, following the idiom of Pyro'sDirichletMultinomialpyro-ppl/pyro#1740If this works well, we may want to add
is_sparseoptions to other discrete distributions used as likelihoods, e.g.GammaPoissonaka negative binomial,DirichletMutlinomial, and maybeMultinomial.Questions for reviewers
is_sparseconstructor arg ok? I slightly prefer this to a separateSparsePoissonsubclass, since I've seen classes with multiple such algorithmic flags and it seems cleaner to avoid a giant n-cube class hierarchy of interacting variants.Tested
SparsePoissontestcc @phylyc