ENH: add exponentiation of a covariance function with a scalar #3852
Conversation
Codecov Report
@@ Coverage Diff @@
## master #3852 +/- ##
==========================================
+ Coverage 90.45% 90.47% +0.02%
==========================================
Files 133 135 +2
Lines 20564 21246 +682
==========================================
+ Hits 18601 19223 +622
- Misses 1963 2023 +60
|
|
Why have the travis tests not started yet? |
pymc3/gp/cov.py
Outdated
| @@ -34,6 +36,7 @@ | |||
| "Coregion", | |||
| "ScaledCov", | |||
| "Kron", | |||
| "Exponentiated", | |||
There was a problem hiding this comment.
Since Exponentiated isn't necessarily user-facing because you added __pow__, it probably shouldn't be included in __all__.
pymc3/gp/cov.py
Outdated
| @@ -117,6 +136,19 @@ def __array_wrap__(self, result): | |||
| raise RuntimeError | |||
|
|
|||
|
|
|||
| class Exponentiated(Covariance): | |||
There was a problem hiding this comment.
Maybe move this down to be near Add and Prod? (very small nitpick, since it's more like a Combination covariance than a base covariance)
There was a problem hiding this comment.
This is great! Maybe you could add a very brief example to the docs so this functionality can be discovered? https://docs.pymc.io/notebooks/GP-MeansAndCovs.html
Tests look perfect.
|
Thanks for the review @bwengals. Noticed that there were no examples exclusively for combinations of kernels so I went ahead and added all of them. I hope they are good to go! |
|
Nice, yeah I support merging now |
|
Thanks @tirthasheshpatel! |
pymc-devs#3852 adds the new feature to the V3.8 changelog while it should go in the V3.9 changelog.
* move change from V3.8 to V3.9 #3852 adds the new feature to the V3.8 changelog while it should go in the V3.9 changelog. * fix small typo * move the change a little down in the changelog
This PR adds functionality to the
pm.gp.covmodule. Now, a covariance function can be exponentiated with a scalar.Changes:
Exponentiatedclass for kernels that are exponentiated with a scalar.__pow__in covariance base class to support exponentiation for all the covariance functions