Skip to content

Changed Categorical to work with multidim p at the logp level reloaded #3386

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

Merged
merged 12 commits into from
Feb 28, 2019
Merged

Changed Categorical to work with multidim p at the logp level reloaded #3386

Changes from 7 commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
5852a5b
Changed Categorical to work with multidim p at the logp level.
lucianopaz Feb 25, 2019
ec77141
Fixed problems with OrderedLogistic.
lucianopaz Feb 25, 2019
0369ffa
Use np.moveaxis instead of transposing. Also added some more tests.
lucianopaz Feb 25, 2019
7e899c6
Changed np.moveaxis to dimshuffle. Removed isinstance conditions that…
lucianopaz Feb 26, 2019
ecbbe09
Merge branch 'master' into multidim_categorical
lucianopaz Feb 26, 2019
beb9cef
Fixed lint error
lucianopaz Feb 27, 2019
080d04c
Resolved conflicts
lucianopaz Feb 27, 2019
8e03412
Force a normalized p in logp
lucianopaz Feb 27, 2019
2dfdd70
Fixed lint error
lucianopaz Feb 28, 2019
2e2c4c5
Merge branch 'master' of https://github.com/pymc-devs/pymc3 into mult…
lucianopaz Feb 28, 2019
1953fde
Edited release notes. Categorical.p can have zeros, and the normaliza…
lucianopaz Feb 28, 2019
88c9e72
Fixed orderedlogistic test logpdf to check for ps>=0 and not just ps>0
lucianopaz Feb 28, 2019
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
8 changes: 3 additions & 5 deletions pymc3/distributions/discrete.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,3 @@
import numbers
import numpy as np
import theano
import theano.tensor as tt
Expand Down Expand Up @@ -739,9 +738,7 @@ def logp(self, value):
# We must only check that the values sum to 1 if p comes from a
# tensor variable, i.e. when p is a step_method proposal. In the other
# cases we normalize ourselves
if not isinstance(p_, (numbers.Number,
np.ndarray,
tt.TensorConstant,
if not isinstance(p_, (tt.TensorConstant,
tt.sharedvar.SharedVariable)):
sumto1 = theano.gradient.zero_grad(
tt.le(abs(tt.sum(p_, axis=-1) - 1), 1e-5))
Expand All @@ -751,7 +748,8 @@ def logp(self, value):
sumto1 = True

if p.ndim > 1:
a = tt.log(np.moveaxis(p, -1, 0)[value_clip])
pattern = (p.ndim - 1,) + tuple(range(p.ndim - 1))
a = tt.log(p.dimshuffle(pattern)[value_clip])
else:
a = tt.log(p[value_clip])

Expand Down