Objective.compute_scalar independent of grid resolution#484
Objective.compute_scalar independent of grid resolution#484
Conversation
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
|
Another thing to possibly consider here: |
|
Can we make a separate PR for all the alphabetization stuff? Doing it here messes up the diff and makes it hard to see what actually changed. |
Yes, but you specifically said at the last dev meeting to just start alphabetizing as we edit files for other reasons... |
yeah that was my bad, I forgot that git sometimes doesn't recognize what changed in the most intuitive way. |
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## master #484 +/- ##
==========================================
+ Coverage 94.43% 94.52% +0.08%
==========================================
Files 79 79
Lines 18497 18417 -80
==========================================
- Hits 17467 17408 -59
+ Misses 1030 1009 -21
|
dpanici
left a comment
dpanici
left a comment
There was a problem hiding this comment.
do we want a changelog entry for this?
4 participants
This PR attempts to resolve most of the problems raised in Issue #482. Most importantly, it makes sure that all objectives with "local" errors now have their
compute_scalarvalues roughly independent of the grid resolution, both when normalized and not.These changes also have
Objective.normalizationnow only used to make the values dimensionless; it removes the need to referenceObjective.dim_fin the compute logic.Also updates
Objective.print_valueto print the min/max/mean of the error instead of just the total, and uses abs error when target is 0, and only prints total error for linear objectives (assumed they're used as exactly satisfied constraints so we don't need lots of details). APrinting should also be more intuitive, printing the actual value of the aspect ratio, volume etc rather than the error.
Resolves #482
Resolves #589