regularization towards initial guess #88
Conversation
nicolearetz
commented
Oct 29, 2025
nicolearetz
commented
- regularization towards an initial guess (defaults to zero if none is provided) + tests
- expansion of operators to larger or smaller basis + tests
- fixed bug in Gram Schmidt algorithm in the method of snapshots
- parameterized polynomial operators of arbitrary order
regularization
shanemcq18
left a comment
shanemcq18
left a comment
There was a problem hiding this comment.
I also need to track down why some of the tests are failing on GitHub, stay tuned.
|
|
||
| For `basics.ipynb`, `inputs.ipynb`, `parametric.ipynb` no additional data generation or download is required. | ||
|
|
||
| For `nested.ipynb`: Download data at URL, or run the script `data_generation_for_nested.py`. No newline at end of file |
src/opinf/lstsq/_tikhonov.py
Outdated
| """Compute the residual of the regularized regression problem.""" | ||
| raise NotImplementedError # pragma: no cover | ||
|
|
||
| def add_initial_guess(self, Ohat): |
There was a problem hiding this comment.
Let's hide these two functions slightly like _add_initial_guess() and _subtract_initial_guess()
|
|
||
| # Execute non-intrusive learning. | ||
| self._extract_operators(self.solver.solve()) | ||
| self._Ohat = self.solver.solve() |
There was a problem hiding this comment.
This is already implemented in the operator_matrix attribute, but not a big deal
| r=self.state_dimension, p=self.polynomial_order | ||
| ) | ||
|
|
||
| def restrict_to_subspace(self, indices_trial, indices_test=None): |
There was a problem hiding this comment.
Do we need a docstring for this?
| ) | ||
| ] | ||
|
|
||
| def extend_to_dimension( |
|
I need to track down an issue with one of the tests, but before that, I was getting some failures on old versions of Python on my local machine that seem to be related to this issue: numpy/numpy#29820. This hopefully shouldn't be a problem for our tests, but it's messing me up locally. It's time to drop python 3.9, so that might help things a bit... Anyway, here's what actually needs to be fixed: |
2 participants
