Граф коммитов

460 Коммитов

Автор SHA1 Сообщение Дата
Keith Battocchi 5732141c77
Correct minor CausalForestDML documenation issue
There is no parameter called "oob" which is referenced in the docs, it's "drate"

Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-03-20 16:32:31 -04:00
Amar Venugopal 27d3101e23
CATE validation - uplift uniform confidence bands (#840)
Add support for multiplier bootstrap uniform confidence band error bars for uplift curves
2024-03-19 12:54:10 -04:00
Keith Battocchi ed4fe33b2b Fix needs_fit logic for model selection with a fixed model
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-28 13:22:24 -05:00
Keith Battocchi 479362aceb Prepare v0.15.0 release
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 15:55:37 -05:00
Keith Battocchi ab572dea46 Enable support for sklearn 1.4
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 15:55:37 -05:00
Keith Battocchi b8a5e2f3c8 Move matplotlib import under guard to avoid error during collection
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 15:55:37 -05:00
Keith Battocchi 329effaf9b Update docstrings
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi 3b35744376 Add model selection tests
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi e4895069b9 Fix RidgeCV model selection
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi d35c340331 Always use generated folds for model selection
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi 639d28cb09 Fix misc. errors flagged by flake8
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi 079d65998b Add model selection documentation
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi 7f2a47525a Extend string options for model selection
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi 4995b76267 Use R2 score for ElasticNet model selection
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi d0d01e6cc2 Enable multiple nuisance fitting passes
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi a886b1a203 Enable pipeline handling during model selection
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi 729e537357 Add needs_fit property
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi a22fd2bb93 Pass discrete outcome to nested model
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi d264cc0b83 Restore _rlearner.py line endings
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-14 12:56:25 -05:00
Keith Battocchi e7c3e3d684 Remove matplotlib backend usage
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-06 11:15:05 -05:00
Keith Battocchi ed4ab1ce2e Update checkout and setup actions to latest versions
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-06 11:15:05 -05:00
Keith Battocchi af7ba5d738 Refactor warning tests to ensure compat with latest pytest
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-06 11:15:05 -05:00
Keith Battocchi 352214adba Remove linear_first_stages from docs and tests
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-06 11:15:05 -05:00
Keith Battocchi f3b0108420 Remove deprecated dynamic module
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-06 11:15:05 -05:00
Keith Battocchi c2582e0079 Add realistic example to docs
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-06 11:15:05 -05:00
Keith Battocchi 0f401dfd34 Make federation opt-in
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2024-02-06 11:15:05 -05:00
fverac ababb7ea30
Improve discreteness handling, allow binary outcomes (#816)
Adds a binary_outcome keyword arg to most estimators, where if True then the outcome nuisance model will be a classifier.

Additionally add constraints to ensure nuisance model discreteness is handled appropriately by the user. 
If a nuisance model has a continuous target but a classifier is passed, then will raise an AttributeError.
Conversely, if a nuisance model has a discrete target but a regressor is passed, then a warning is issued.
2024-01-12 15:31:52 -05:00
Amar Venugopal 67eef1e191
CATE uplift validation methods (#836)
Added additional functionality to the DRTester validation class to include AUTOC validation metric, with associated inference methods. Also included ability to plot uplift curve methods (both QINI and TOC curves) and cleaned up handling of multiple treatments.
2024-01-12 12:00:25 -05:00
Christine P. Chai 7793184ed3
Corrected a typo in dml.rst (#831)
Corrected a typo in dml.rst: efffect -> effect

Signed-off-by: Christine P. Chai <star1327p@gmail.com>
2023-12-01 07:36:29 -05:00
Keith Battocchi d7c593c2e8 Prepare v0.15.0b1 release
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-11-11 17:23:34 -05:00
Keith Battocchi 7a3202ee4c Update LKG
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-11-11 17:23:34 -05:00
Keith Battocchi 59861d3055 Remove unnecessary extras from tests
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-11-11 17:23:34 -05:00
Keith Battocchi 600428f288 Ensure tests can be collected without optional dependencies
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-11-11 17:23:34 -05:00
Keith Battocchi 720ac4365e Updated supported shap versions
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-11-11 17:23:34 -05:00
Keith Battocchi 05ef0bb847 Update supported python versions
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-11-11 17:23:34 -05:00
Keith Battocchi e335d152f0
Enable model selection for first stage models (#808)
Enables model selection for _OrthoLearner first stage models, and changes all concrete subclasses to allow selection between linear and random forest models for all first stage models by default.
---------

Signed-off-by: AnthonyCampbell208 <78286293+AnthonyCampbell208@users.noreply.github.com>
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
Co-authored-by: AnthonyCampbell208 <78286293+AnthonyCampbell208@users.noreply.github.com>
Co-authored-by: ShrutiRM97 <98553136+ShrutiRM97@users.noreply.github.com>
Co-authored-by: CooperGibbs <coopergibbs@Coopers-MacBook-Pro.local>
2023-11-11 19:57:59 +00:00
Amar Venugopal a10e42d438
CATE validation suite (#777)
Add new CATE validation class, DRTester, to perform doubly-robust based CATE validation methods. Implemented methods include BLP, calibration, and QINI coefficient.

---------

Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
Co-authored-by: Evan Flack <evanjflack@gmail.com>
Co-authored-by: Keith Battocchi <kebatt@microsoft.com>
2023-11-08 20:39:21 +00:00
Kevin Gao f170babe07
Kgao/federal learning (#818)
Add the FederatedEstimator class for federated learning of linear CATE models

Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
Signed-off-by: kgao <kevin.leo.gao@gmail.com>
Co-authored-by: Keith Battocchi <kebatt@microsoft.com>
2023-11-02 02:15:04 +00:00
Vishal Verma 01899a8957
Scaling ortholearners using Ray (#800)
Added Implementation of ray-based distributed parallelization to crossfit.

---------

Signed-off-by: Vishal Verma <vishalmverma27@gmail.com>
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
Co-authored-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-27 18:17:14 -04:00
Keith Battocchi 5423183471 Remove legacy assertWarns hack
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi 95f6dfdc4a Save notebook outputs during CI
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi 6c0848f1e6 Make minor CI improvements
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi 53e70081b4 Allow newer shap, matlab, and seaborn versions
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi 499ef18793 Make changes to support dowhy 0.10.1 in tests
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi 9a0dbdbad8 Update __init__.py to reflect current structure
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi cac0108103 Ensure groups work with DRIV, DMLIV
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi 6df8935125 Enable sklearn 1.3
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi 21f0b83e81 Support direct covariance fitting in DRIV
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
Keith Battocchi 105e7261bb Drop support for sklearn<1.0
Signed-off-by: Keith Battocchi <kebatt@microsoft.com>
2023-10-25 01:38:49 -04:00
fverac 25c3b3b82a
add arg to allow missing values in W and sometimes X (#791)
* enable nans in W

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* linting

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* add tests for ests that handle missing in W

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* allow missing in X for some ortholearner subclasses

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* refactor keyword arg to be bool only, add more tests

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* linting

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* enable missing for metalearners and orf, fix dowhywrapped discretetreat dmlorf

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* update arg name to allow_missing, add docstrings

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* add warning when missing values detected

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* dummy commit

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

* dummy commit revert

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>

---------

Signed-off-by: Fabio Vera <fabiovera@microsoft.com>
2023-09-29 17:37:11 -04:00