feat: optimize stretch using a Hessian matrix#204
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #204 +/- ##
==========================================
+ Coverage 77.50% 82.00% +4.50%
==========================================
Files 3 3
Lines 40 50 +10
==========================================
+ Hits 31 41 +10
Misses 9 9
🚀 New features to boost your workflow:
|
| gra = gra.flatten() | ||
| return fun, gra | ||
|
|
||
| def hessian(stretch_vec): |
There was a problem hiding this comment.
do we need a test for this?
There was a problem hiding this comment.
Good idea. I have added a test checking for some of the expected properties (right shape, symmetric, correct ordering/cross-coupling).
| return fun, gra | ||
|
|
||
| def _regularize_function_hessian(self, stretch): | ||
| residuals, d_stretch_comps, dd_stretch_comps = ( |
There was a problem hiding this comment.
this could use some dicumentation and also maybe a more descriptive variable name. Stretch is a verb usually and I would expect a noun here.
We were pretty lazy about documenting code in the early part of the project because we wanted it to get going from a low place, but we are at the point of the project where we can take the time to make the code more maintainable moving forward.
2 participants
This is another PR bringing us closer to parity with MATLAB. Proper use of a Hessian was key to the speed and quality of convergence they were able to obtain. I had tried this a few times before, but it never worked even though the Hessian had the right shape. By properly regularizing, it now works.