HalfCauchy, Gamma, Weibull and LogNormal moments #5148
Conversation
lucianopaz
force-pushed
the
lognormal_moments
branch
from
4d630c4 to
e809a0a
Compare
lucianopaz
changed the title
Codecov Report
@@ Coverage Diff @@
## main #5148 +/- ##
==========================================
+ Coverage 77.82% 77.86% +0.03%
==========================================
Files 88 88
Lines 14145 14165 +20
==========================================
+ Hits 11008 11029 +21
+ Misses 3137 3136 -1
|
pymc/distributions/continuous.py
Outdated
| @@ -2100,6 +2106,12 @@ def dist(cls, beta, *args, **kwargs): | |||
| assert_negative_support(beta, "beta", "HalfCauchy") | |||
| return super().dist([0.0, beta], **kwargs) | |||
|
|
|||
| def get_moment(rv, size, loc, beta): | |||
| mean, _ = at.broadcast_arrays(at.as_tensor(0.0, dtype=rv.dtype), beta) | |||
There was a problem hiding this comment.
Does something like this also work?
| mean, _ = at.broadcast_arrays(at.as_tensor(0.0, dtype=rv.dtype), beta) | |
| mean = at.zeros_like(beta) |
There was a problem hiding this comment.
Is zero a good point to start sampling from this distribution?
There was a problem hiding this comment.
Looking at V3 I think the median (beta) takes precedence over the mode (0), so we were probably using the former to initialize halfcauchys
There was a problem hiding this comment.
Oh, I'll use the median instead then. Maybe we should list the order of preference for the moments in #5078
|
@lucianopaz seems like pre-commit is failing with import sorting |
lucianopaz
force-pushed
the
lognormal_moments
branch
from
e809a0a to
f5407b1
Compare
Yeah I messed up the pre commit while rebasing the PR. But now everything should be fixed |
|
👍 |
Add moments and tests for the below distributions as part of #5078