Dev/master#197
Merged
Merged
Conversation
Dev/tutorials
# Conflicts: # hypex/aa.py
Dev/custom weights
Dev/custom weights
…tom_weights # Conflicts: # hypex/ui/matching.py
TonyKatkov89
added a commit
that referenced
this pull request
Jun 18, 2026
* Base executors in the executors family scheme. * Add analyzers in executors_famaly.dot * Start abstract comporators * Abstract comporators are finished * Comporators are finished * Add Data, Encoders and Extensions * Add forks * Micro refactor and new schemes: * ml * operators * reporters * New architecture sceme * Update scheme * Написали от начала до экспериментов * Add Reporters * Add Extensions * Rework part 11 * Add toc * Add part of presentation * Add part of presentation * Dev/master (#197) * aan added * best split assembled correctly * tutorail updated, test added * mantching n-neighbours added * add cuped * custom_weights added * matching fixed * one feature fix * quality result fix * const_groups fix * ab multitest fixed * add min_sample_size * add feature importances * tox fixed --------- Co-authored-by: anastasiiafed24 <fedorovanasty24@gmail.com> Co-authored-by: anastasiiafed24 <144920163+anastasiiafed24@users.noreply.github.com> Co-authored-by: Alsherov Ruslan <114132014+anathema-git@users.noreply.github.com> Co-authored-by: Альшеров Руслан <ruslan-alsherov@yandex.ru> Co-authored-by: Dasha Vigovskaya <vigosya@Dashas-MacBook-Air.local> Co-authored-by: Vigovskaya Daria <78546846+Vigosya@users.noreply.github.com> Co-authored-by: Альшеров Руслан <uslan-alsherov@yandex.ru> * Ks test in ab (#216) * ks-test for ab-test added * gitignore updated * version update * scipy requirements updated * scipy requirements updated * Cupac fix (#226) * fix: CUPAC validation, KSTest pass display, and max_lag error handling - Fix KSTest (and other stat tests) always showing "OK" due to `is True` identity check failing for numpy.bool_ values; use `bool(v)` instead - Add clear ValueError when a CUPAC target has no lag periods defined, advising to use PreTargetRole(lag=N) - Improve CUPAC validation error message to mention uninstalled packages * ttest equal_vars changed to False by default * ttest equal_vars changed to False by default * version and readme updated * version and readme updated * documentation update * deprecation warning added * tutorails and ok/not_ok assignment fixed * Update ci.yml (#227) * Fix CUPED theta to Cov(X,Y)/Var(X) The CUPED adjustment used theta = Cov(X,Y) / (std_x * std_y), which equals the correlation coefficient rho and is only correct when std_x == std_y. For covariates and targets on different scales this left variance on the table (or, when std_x >> std_y, increased variance). Use the optimal theta = Cov(X,Y) / Var(X) so the adjustment reaches the rho**2 reduction bound. cov_xy and var_x are now both population moments so the 1/N factors cancel exactly, and the zero-variance guard checks the denominator (var_x). Add tests/test_cuped.py pinning rho**2 variance reduction across differing X/Y scales, the theta = Cov/Var identity, and the constant-covariate no-op. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * Apply CUPED theta-scaling to CUPAC adjustment CUPAC removed the model prediction from the target directly (y - (pred - E[pred])), implicitly assuming theta = 1. That is only optimal when Cov(pred, y) == Var(pred); for cross-period and regularized predictions the scale differs, so it left variance on the table (and reported variance_reduction_real below variance_reduction_cv). Residualize with the optimal CUPED coefficient theta = Cov(pred, y)/Var(pred) in both CUPAC paths: - CV/model-selection (extensions/cupac.py): pool out-of-fold predictions and apply a single theta-residualization, matching cross-fitted CUPED. - Final adjustment (ml/cupac.py): scale the explained variation by theta. _cuped_theta guards a (near-)constant prediction with a scale-relative threshold so theta=0 (no-op) instead of a ratio of floating-point noise. Add tests/test_cupac.py: theta = Cov/Var identity, theta-residualization never worse than theta=1 (strictly better under scale mismatch), constant prediction no-op, and an end-to-end ABTest CUPAC variance-reduction check. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * Tune CUPAC default Ridge/Lasso hyperparameters Set Ridge(alpha=0.5) and Lasso(alpha=0.01, max_iter=10000) as the default CUPAC regressors instead of sklearn defaults (alpha=1.0). The default Lasso(alpha=1.0) under-fits and loses ~14pp of variance reduction, which could flip CUPAC model selection toward Linear even when a regularized model fits better. With these defaults the per-model variance reductions and selected best model match the reference CUPAC implementation exactly. --------- Co-authored-by: Dmatryus <dmatryus.sqrt49@yandex.ru> Co-authored-by: dmbulychev <dmbulychev@avito.ru> Co-authored-by: anastasiiafed24 <fedorovanasty24@gmail.com> Co-authored-by: anastasiiafed24 <144920163+anastasiiafed24@users.noreply.github.com> Co-authored-by: Alsherov Ruslan <114132014+anathema-git@users.noreply.github.com> Co-authored-by: Альшеров Руслан <ruslan-alsherov@yandex.ru> Co-authored-by: Dasha Vigovskaya <vigosya@Dashas-MacBook-Air.local> Co-authored-by: Vigovskaya Daria <78546846+Vigosya@users.noreply.github.com> Co-authored-by: Альшеров Руслан <uslan-alsherov@yandex.ru> Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
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
5 participants
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.