Feature prior #130
Feature prior #130
Conversation
Codecov Report
@@ Coverage Diff @@
## develop #130 +/- ##
===========================================
+ Coverage 67.3% 68.44% +1.13%
===========================================
Files 11 11
Lines 1040 1090 +50
Branches 241 256 +15
===========================================
+ Hits 700 746 +46
- Misses 305 307 +2
- Partials 35 37 +2
Continue to review full report at Codecov.
|
paulstapor
requested review from
yannikschaelte,
LeonardSchmiester and
jvanhoefer
and removed request for
dweindl
| sp = scale(np.random.normal(loc=p_params[0], scale=p_params[1], | ||
| size=(n_starts,))) | ||
|
|
||
| elif p_type == 'logNormal': |
There was a problem hiding this comment.
We need to decide on a consistent parametrization of the distributions.
I understood from our discussion (might be my misunderstanding...), that we agreed on another parametrization, but I prefer your parametrization here. What do you think @LeonardSchmiester? :)
There was a problem hiding this comment.
I thought that was the way we agreed on it: Introducing log-normal as an exponentiated normal distribution on the linear parameter space? What should be different?
There was a problem hiding this comment.
I thought that we parametrized it as it is parametrized in scipy. But as I said, I prefer your version and just wanted to make sure, that Leonard implements the same parametrization. I will adapt my code accordingly.
There was a problem hiding this comment.
Ah...
Yes, scipy does it differently...
Yes, but we do everything via numpy here, so I would also prefer not to get involved with yet another package which has redundant functionality, but uses a different definition...
|
@paulstapor : Can you please include a section in the documentation on the (meaning of the) supported values for the respective columns? |
Indeed, that's still missing. Will do that before merging! |
…ctivePriorType and Parameters
|
Okay, I updated the documentation, also about the additional columns we're actually still discussing about (initializationPriorType, initializationPriorParameters, objectivePriorType, and objectivePriorParameters). This is yet to come, but fairly simple to implement. I think/hope, it was safe to include this in the "Extensions" part of the documentation... |
|
Can we agree on that this fixes now #17 ? |
Okay for me. |
|
May/Can/Shall I merge? |
Both things (this PR and the Issue) are PEtab things... |
Yeah, sorry. Just ignore my comment 🤦♂ |
* added an initial parameter sampling functionality * add tests for startpoint sampling * added documentation for the additional columns intialization and objectivePriorType and Parameters
Added some functionality to sample starting points according to priors, added files for a test (Fixes #17 )