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.
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
MGP #60
MGP #60
Conversation
Interfacing still needs cleaning up
Interfacing still needs cleaning up
Interfacing still needs cleaning up
Awesome, thanks. I see that the Travis tests aren't triggered because of an external PR. In init.py you removed the objective import by accident I think, as well as forgot to remove testmgp.py But we will fix that once I moved the branch and created a new PR |
This is not to be included in the 0.1 release, preferable after tf 1.3 actual release I suggest to fix things first, then create an in-house branch as @nknudde has no rights to push after that. Some things to consider:
|
Codecov Report
@@ Coverage Diff @@
## master #60 +/- ##
==========================================
+ Coverage 99.78% 99.79% +0.01%
==========================================
Files 16 17 +1
Lines 928 987 +59
==========================================
+ Hits 926 985 +59
Misses 2 2
Continue to review full report at Codecov.
|
return self.build_predict(fmean, fvar, theta) | ||
|
||
@AutoFlow((float_type, [None, None]), (float_type, [None, None])) | ||
def predict_density(self, Xnew, Ynew): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Still missing a test for predict density :)
Small general guideline: lets avoid rebases. For this PR I'm ok with it, and we won't start from scratch. Rebases are actually very good but for normal circumstances I think they make the history confusing. its a very good tool for complex merge scenario with an enormous amount of conflicts expected as it allows you to resolve conflicts as they occur, commit by commit (think of a fork which has gone its own way for a while and then is merged). For normal scenarios a merge is preferred. |
@@ -8,7 +8,7 @@ python: | |||
cache: pip | |||
install: | |||
- pip install -U pip wheel | |||
- pip install tensorflow==1.0.1 | |||
- pip install tensorflow==1.3.0rc0 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
can we later specify this without a version number so we just always test against the latest stable? Which is the goal anyways, or rather work with the latest stable GPflow which hopefully follows the latest stable TF closely.
I'm also against rebases (at least I am now, some bad experiences). I thought it might be necessary to fix the commit history a bit as i thought it was corrupted due to rebranching, but it actually looks okay ( though there are several commits with the same name) |
In this pull request I implemented the Approximatively Marginalised GP as described as indicated in issue #39 . It currently supports multi-output GPs. A notebook and some tests are included.