Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Improve STLSQ #36

Closed
briandesilva opened this issue Jan 31, 2020 · 3 comments
Closed

Improve STLSQ #36

briandesilva opened this issue Jan 31, 2020 · 3 comments
Labels
enhancement New feature or request priority:low

Comments

@briandesilva
Copy link
Member

Setting the alpha (L2 regularization) parameter for the STLSQ optimizer improves the condition number of the least-squares problems that are solved, leading to better overall results when features are close to linearly dependent. However, it adds some bias to the learned coefficients (toward 0). We should consider modifying STLSQ (and maybe some of our other methods) by changing it to a two-step procedure when alpha is positive:

  1. Use the existing algorithm with regularization to identify the support
  2. Perform one last unregularized least-squares fit using only the features/terms identified in step 1.
@briandesilva briandesilva added enhancement New feature or request priority:low labels Jan 31, 2020
@Ohjeah
Copy link
Collaborator

Ohjeah commented Jan 31, 2020

This was already a feature in the upstream repo sparsereg, see https://github.com/Ohjeah/sparsereg/blob/master/sparsereg/model/base.py#L219 and can be cherry-picked.

@Ohjeah
Copy link
Collaborator

Ohjeah commented Jan 31, 2020

This needs to be fixed for the other optimizers as well.

@briandesilva
Copy link
Member Author

Fixed in #39

jpcurbelo pushed a commit to jpcurbelo/pysindy_fork that referenced this issue May 9, 2024
@gauchm
CustomLSTM tutorial (dynamicslab#65)
c271d04 
Add CustomLSTM tutorial (part of dynamicslab#36 and dynamicslab#64)
jpcurbelo pushed a commit to jpcurbelo/pysindy_fork that referenced this issue May 9, 2024
@kratzert @gauchm
added finetuning notebook dynamicslab#36, fixed missing links in docs (…
0c38c3d 
…dynamicslab#69)

* added finetuning notebook dynamicslab#36, fixed missing links in docs

* Apply suggestions from code review

Co-authored-by: Martin Gauch <15731649+gauchm@users.noreply.github.com>

Co-authored-by: Martin Gauch <15731649+gauchm@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
enhancement New feature or request priority:low
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@Ohjeah @briandesilva