Upgrade examples - #31
Merged
Merged
Conversation
Example 01 (basic_fitting):
- Regenerate synthetic data over a longer pre-perturbation baseline
(time -20..100, 481 slices) so GLP.m is well-constrained and can vary
instead of being fixed; add models_{energy,time}_truth.yaml and a
generate_data.ipynb that writes the CSVs with a stable %.6e format.
- Update example.ipynb: absolute-time baseline window (-20..-10, clear of
the IRF-smeared onset), matching fit limits, and explanatory text.
Library:
- Slice-by-slice serial path uses a tqdm progress bar instead of a raw
carriage-return print, which flooded nbconvert/Jupyter output. Plain
tqdm (not tqdm.auto) to avoid the ipywidgets dependency.
- fit_slice_by_slice / fit_2d now display the data/fit/residual maps
inline when show_output>=1 even without auto_export, mirroring
fit_baseline; CSV/PNG writes stay gated behind auto_export via a new
save_files flag on the legacy export helpers.
- Standardize results_to_df save_df on the {-2,-1,0,1} finalize-plot
convention (default -2; 0 now means display-only).
Docs / examples:
- Add example 12_uncertainty_mcmc and wire it into the docs index,
quickstart, and example READMEs.
Tests:
- Cover the show_output>=1 + auto_export=False display path for
fit_slice_by_slice and fit_2d (maps shown, no files written).
Move to synthetic data Do not reject convolution components if the model is a single component model (use case is a global IRF in multi-cycle dynamics models)
Separated data generation into its own notebook similar to the other examples
context manager to correctly spawn parallel processes for both noteboook and standard python callers
- rename fit_wrapper's `sigmas` -> `ci_sigmas`, disambiguating the CI/MCMC confidence levels from the data-noise sigma (File.set_sigma / sigma_data) - make the MCMC noise-scale nuisance parameter configurable via MC(sigma_ini/sigma_min/sigma_max) instead of a hardcoded log(0.1)/[0.001,2] - add MCMCResult (utils/lmfit) and the live File accessors get_correlations / get_conf_intervals / get_mcmc (+ _result_model), so callers read fit outputs through the public API instead of indexing result[1..4] - fix fit_wrapper mutating par_fin.params in place: deepcopy before adding __lnsigma so the MCMC nuisance parameter no longer leaks into the stored leastsq result (result[1]) and its downstream consumers - tests: MC sigma-knob validation/reach, the __lnsigma-leak regression, and the three new accessors
- update docs to clarify 11 and 12 are based on other examples - consistently use file from archive instead of session file
update quickstart and index docs as well
- output only parameters that are varied - this aligns with lmfit.conf_interval bahavior (and others)
- rename 20_fit_each_separately -> 20_multi_file_independent_fit (clearer; matches the notebook H1). Update all references: READMEs, docs/quickstart, docs/examples/index.rst, examples_upgrade.md, CHANGELOG, TODO, .gitignore, and the notebook-11 cross-link. - fix systematic baseline bias: the SD~3 IRF onset at t0=0 leaked the peak shift back into the baseline window (t<=-4, ~1.3 sigma), biasing the static ground-state params in proportion to each file's shift amplitude. Pull the window to t<=-8.5 (time_stop 12 -> 3) so static params recover truth and the residual dipole clears (data_1 chi2_red_raw 0.10 -> 0.05). - correct compare_models prose: the survey is not pre-ranked (sort it yourself) and carries no per-parameter columns. - reword "bridge to 2x" -> "bridge to shared-parameter fitting". - gitignore the generated batch_export/ export tree.
- rename 21_project_level_shared_fit -> 21_multi_file_shared_fit (parallel to 20). Update all references - extend basline to earlier times and increase noise level similar to example 01 (chi2_red_raw ~1.0 (was an over-polished ~0.05)) - add Section 4 (shared-vs-per-file parameter table checked against truth) Note: regenerating the shared dataset also updates example 20's notebook (data path, time limits, baseline window) and the 6 CSVs.
test ran skill on existing examples and fixed the issues it surfaced
housekeeping before PR
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
No description provided.