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.

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

Resolving some test problem consistency and apparent UKI instability #294

Merged
merged 1 commit into from
Jun 14, 2023

Conversation

odunbar
Copy link
Collaborator

@odunbar odunbar commented Jun 8, 2023

Purpose

Closes #287

Content

  • more consistent plotting between tests

  • adjustments to the test problems

  • removed inappropriate Localization tests in EnsembleKalmanProcess/runtests.jl examples, (they are already properly tested in Localizers/runtests.jl ) - These caused blow up without disturbing the ensemble mean, therefore added some spread-checking too to ensure the tr(ensemble cov) does not increase too much

  • test UKI with nonlinear problems, not just linear ones

  • Added prior Tikhonov regularization for UKI, modified test checks to account for the regularization.

  • Changed the default update_freq in UKI from 1 to 0. I.e. uses prior cov not current cov by default.

  • Minor type change in UKI constructor to allow arrays to be converted to float type internally

  • added note to docs RE this update


  • I have read and checked the items on the review checklist.

UPDATED Example:

To reproduce run julia --project; ENV["CES_TEST_PLOT_OUTPUT"] = true; ]; test and look for the figures XYZ_test_4.png in the runtest directory.

Problem: x in R^2, and y = (||x-x*||_1,||x-x*||_2, ||x-x*||_infty ) + N(0,0.01), 100 iterations of methods:
Prior bounded by [-2,2] in x axis

EKI & EKS

Tikhonov UKI quad & simplex

@odunbar odunbar force-pushed the orad/test_problems_and_uki branch 2 times, most recently from 9d57eb4 to e001485 Compare June 14, 2023 16:08
plot consistently

rm verbose

removed the inappropriate localizations in EKP/runtest

notation

formatting

UKI tested on nonlinear problems too

format

using \alpha to get convergence with UKI

100 iterations

update UKI

update default hyperparameters

format and remove hard-code flag

new 3D output test

format and all test pass

add abs in nonlinear inverse problem test; update UKI default

format

docs note for UKI impose_prior setting

docs update

default prior to initial cov/mean and make constructor a little more flexible

tests for default prior to init

codecov

format...
Copy link
Contributor

@Zhengyu-Huang Zhengyu-Huang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this PR looks great! Thank you for the great work!

@odunbar
Copy link
Collaborator Author

odunbar commented Jun 14, 2023

bors r+

@bors
Copy link
Contributor

bors bot commented Jun 14, 2023

Build succeeded!

The publicly hosted instance of bors-ng is deprecated and will go away soon.

If you want to self-host your own instance, instructions are here.
For more help, visit the forum.

If you want to switch to GitHub's built-in merge queue, visit their help page.

@bors bors bot merged commit b18af99 into main Jun 14, 2023
11 checks passed
@bors bors bot deleted the orad/test_problems_and_uki branch June 14, 2023 22:41
@odunbar odunbar changed the title [WIP] Resolving some test problem consistency and apparent UKI instability Resolving some test problem consistency and apparent UKI instability Jul 5, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

UKI improperly tested - and other test issues...
2 participants