ENH(sr3): refactor threshold and thresholder property #548
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
- move to private function - rename lam to regularization_weight - remove cad - add docstring - add validation
himkwtn
force-pushed
the
refactor-sr3
branch
from
e6de7cc to
e49304d
Compare
|
Thanks Watcharin! We can talk about this today, but this PR should target |
Comment on lines
+156
to
+138
| if regularizer == "l1": | ||
| cost = cp.sum_squares(x[:, i] - x @ xi) + cp.sum( | ||
| cp.multiply(lam, cp.abs(xi)) | ||
| ) | ||
| elif regularizer == "weighted_l1": | ||
| cost = cp.sum_squares(x[:, i] - x @ xi) + cp.sum( | ||
| cp.multiply(lam[:, i], cp.abs(xi)) | ||
| ) | ||
| elif regularizer == "l2": | ||
| cost = cp.sum_squares(x[:, i] - x @ xi) + cp.sum( | ||
| cp.multiply(lam, xi**2) | ||
| ) | ||
| elif regularizer == "weighted_l2": | ||
| cost = cp.sum_squares(x[:, i] - x @ xi) + cp.sum( | ||
| cp.multiply(lam[:, i], xi**2) | ||
| ) |
There was a problem hiding this comment.
Can we find a way to use _calculate_penalty?
There was a problem hiding this comment.
The logic is slightly different from ConstrainedSR3. I'm not sure if we can combine them.
There was a problem hiding this comment.
If it doesn't work, override the superclass _calculate_penalty()
- rename to reg_weight and regularizer - update all related test cases - fix all child classes
- reg_weight to reg_weight_lam - nu to relax_coeff_nu
himkwtn
force-pushed
the
refactor-sr3
branch
from
80bf1e1 to
5e6b777
Compare
# Conflicts: # examples/1_feature_overview/example.ipynb # examples/1_feature_overview/example.py # pysindy/optimizers/constrained_sr3.py # pysindy/optimizers/sr3.py # pysindy/optimizers/stable_linear_sr3.py # test/test_optimizers.py
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #548 +/- ##
==========================================
+ Coverage 94.23% 94.64% +0.40%
==========================================
Files 37 37
Lines 4045 4018 -27
==========================================
- Hits 3812 3803 -9
+ Misses 233 215 -18 ☔ View full report in Codecov by Sentry. |
himkwtn
commented
Sep 11, 2024
Comment on lines
+156
to
+138
| if regularizer == "l1": | ||
| cost = cp.sum_squares(x[:, i] - x @ xi) + cp.sum( | ||
| cp.multiply(lam, cp.abs(xi)) | ||
| ) | ||
| elif regularizer == "weighted_l1": | ||
| cost = cp.sum_squares(x[:, i] - x @ xi) + cp.sum( | ||
| cp.multiply(lam[:, i], cp.abs(xi)) | ||
| ) | ||
| elif regularizer == "l2": | ||
| cost = cp.sum_squares(x[:, i] - x @ xi) + cp.sum( | ||
| cp.multiply(lam, xi**2) | ||
| ) | ||
| elif regularizer == "weighted_l2": | ||
| cost = cp.sum_squares(x[:, i] - x @ xi) + cp.sum( | ||
| cp.multiply(lam[:, i], xi**2) | ||
| ) |
There was a problem hiding this comment.
If it doesn't work, override the superclass _calculate_penalty()
examples/8_trapping_sindy_examples/von_karman_trapping_extended.py
Outdated
Show resolved
Hide resolved
examples/8_trapping_sindy_examples/von_karman_trapping_extended.py
Outdated
Show resolved
Hide resolved
examples/8_trapping_sindy_examples/von_karman_trapping_extended.py
Outdated
Show resolved
Hide resolved
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Threshold parameter is specific to l0 norm. This PR aims to generalize SR3 to other norm by moving threshold calculation to a separate function and using generic regularization weight instead.