Closed kyr-pol closed 5 years ago
Merging #16 into master will decrease coverage by
1.25%
. The diff coverage is90.76%
.
@@ Coverage Diff @@
## master #16 +/- ##
==========================================
- Coverage 96.37% 95.12% -1.26%
==========================================
Files 7 7
Lines 276 328 +52
==========================================
+ Hits 266 312 +46
- Misses 10 16 +6
Impacted Files | Coverage Δ | |
---|---|---|
pilco/models/mgpr.py | 100% <100%> (ø) |
:arrow_up: |
pilco/models/pilco.py | 92.94% <100%> (+2.03%) |
:arrow_up: |
pilco/controllers.py | 88.73% <45.45%> (-7.94%) |
:arrow_down: |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update 90998ac...daf44da. Read the comment docs.
@kyr-pol many thanks for this. I have cleaned up a bit the code and I think that this can now be merged.
Let me know if you want to change something before merging.
Slowly merging changes from the more_envs branch to master.
In this pull request, I added the restarts, of the controller and the model, along with the modification we had to do to the optimizer due to the graph expanding (see discussion in issue #13). Units tests are passing, but I haven't written new tests for the restarts. Since
there is no Matlab equivalent to test against
what we could do is perform a number of restarts and make sure that the objective is always getting better or equal to the starting value (as it should, because when the restart fails the previous values are restored ). It's statistical, so to make sure we should perform more than one, but one the other hand we don't want the unit tests to take too long.
What needs to be added also is a restart function for the smgpr.py, because the one used for the standard mgpr.py does not work.