CHORE: Drop support for Python <3.9 and NumPy <1.22

.github/workflows/test.yml

@@ -15,15 +15,9 @@ jobs:
     strategy:
       matrix:
         include:
-          - os: ubuntu-latest
-            python-version: "3.7"
-            numpy-version: 1.21.4
-          - os: ubuntu-latest
-            python-version: "3.8"
-            numpy-version: 1.21.4
           - os: ubuntu-latest
             python-version: "3.9"
-            numpy-version: 1.21.4
+            numpy-version: 1.22.4
           - os: ubuntu-latest
             python-version: "3.10"
             numpy-version: 1.22.4

README.rst

@@ -79,8 +79,8 @@ Implemented methods in **MAPIE** respect three fundamental pillars:
 🔗 Requirements
 ===============
 
-- **MAPIE** runs on Python 3.7+.
-- **MAPIE** stands on the shoulders of giants. Its only internal dependencies are `scikit-learn <https://scikit-learn.org/stable/>`_ and `numpy=>1.21 <https://numpy.org/>`_.
+- **MAPIE** runs on Python 3.9+.
+- **MAPIE** stands on the shoulders of giants. Its only internal dependencies are `scikit-learn <https://scikit-learn.org/stable/>`_ and `numpy=>1.22 <https://numpy.org/>`_.
 
 
 🛠 Installation

doc/v1_migration_guide.rst

@@ -5,7 +5,14 @@ MAPIE v1 introduces several updates, enhancements, and structural changes that s
 
 This guide outlines the differences between MAPIE v0.x and MAPIE v1 and provides instructions for migrating your code to the new API.
 
-1. Overview of class restructuring
+1. Python and NumPy versions support
+-------------------------------------------------
+
+We now support Python >= 3.9 (formerly >=3.7), and NumPy >= 1.22 (formerly >=1.21).
+
+MAPIE v1 may run with Python < 3.9, but we do not recommend it.
+
+2. Class restructuring
 -----------------------------------
 
 MAPIE v1 breaks down the ``MapieRegressor`` and ``MapieClassifier`` classes into 5 classes, each dedicated to a particular conformal prediction technique. ``MapieQuantileRegressor`` has also been revamped, and renamed ``ConformalizedQuantileRegressor``.
@@ -48,7 +55,7 @@ The ``cv`` parameter is key to understand what new class to use in the v1 API:
 
 For more details regarding the difference between split and cross conformal types, see :doc:`split_cross_conformal`
 
-2. Method changes
+3. Method changes
 -----------------
 
 In MAPIE v1, the conformal prediction workflow is more streamlined and modular, with distinct methods for training, conformalization (named calibration in the scientific literature), and prediction. The conformalization process in v1 consists of four steps.
@@ -91,7 +98,7 @@ The ``.predict()`` method now focuses solely on producing point predictions.
 
 
 
-3. Parameters change
+4. Parameters change
 ------------------------
 
 ``alpha``
@@ -182,7 +189,7 @@ No more parameters with incorrect ``None`` defaults.
 - **v1**: All parameters now have explicit defaults.
 
 
-4. Migration example: MAPIE v0.x to MAPIE v1
+5. Migration examples: MAPIE v0.x to MAPIE v1
 ----------------------------------------------------------------------------------------
 
 Below is a side-by-side example of code in MAPIE v0.x and its equivalent in MAPIE v1

examples/calibration/1-quickstart/plot_calibration_hypothesis_testing.py

@@ -29,7 +29,7 @@
 from matplotlib import pyplot as plt
 from sklearn.utils import check_random_state
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.metrics import (cumulative_differences, kolmogorov_smirnov_p_value,
                            length_scale)
 

examples/calibration/2-advanced-analysis/plot_asymptotic_convergence_of_p_values.py

@@ -39,7 +39,7 @@
 from matplotlib import pyplot as plt
 from sklearn.utils import check_random_state
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.metrics import (kolmogorov_smirnov_p_value, kuiper_p_value,
                            spiegelhalter_p_value)
 

examples/classification/1-quickstart/plot_comp_methods_on_2d_dataset.py

@@ -53,7 +53,7 @@
 from sklearn.model_selection import train_test_split
 from sklearn.naive_bayes import GaussianNB
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.classification import MapieClassifier
 from mapie.metrics import (classification_coverage_score,
                            classification_mean_width_score)

examples/classification/4-tutorials/plot_crossconformal.py

@@ -35,7 +35,7 @@
 from sklearn.naive_bayes import GaussianNB
 from typing_extensions import TypedDict
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.classification import MapieClassifier
 from mapie.metrics import (classification_coverage_score,
                            classification_mean_width_score)

examples/classification/4-tutorials/plot_main-tutorial-binary-classification.py

@@ -33,7 +33,7 @@
 from sklearn.model_selection import train_test_split
 from sklearn.naive_bayes import GaussianNB
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.classification import MapieClassifier
 from mapie.metrics import (classification_coverage_score,
                            classification_mean_width_score)

examples/regression/1-quickstart/plot_heteroscedastic_1d_data.py

@@ -25,7 +25,7 @@
 from sklearn.preprocessing import PolynomialFeatures
 from sklearn.model_selection import train_test_split
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie_v1.regression import (
     CrossConformalRegressor,
     JackknifeAfterBootstrapRegressor,

examples/regression/1-quickstart/plot_homoscedastic_1d_data.py

@@ -23,7 +23,7 @@
 from sklearn.model_selection import train_test_split
 
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie_v1.regression import (
     CrossConformalRegressor,
     JackknifeAfterBootstrapRegressor,

examples/regression/1-quickstart/plot_prefit.py

@@ -26,7 +26,7 @@
 from matplotlib import pyplot as plt
 from sklearn.neural_network import MLPRegressor
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.metrics import regression_coverage_score
 from mapie_v1.regression import SplitConformalRegressor, ConformalizedQuantileRegressor
 from mapie_v1.utils import train_conformalize_test_split

examples/regression/2-advanced-analysis/plot_both_uncertainties.py

@@ -19,7 +19,7 @@
 from sklearn.preprocessing import PolynomialFeatures
 from sklearn.model_selection import train_test_split
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie_v1.regression import (
     CrossConformalRegressor,
     ConformalizedQuantileRegressor,

examples/regression/2-advanced-analysis/plot_conditional_coverage.py

@@ -26,7 +26,7 @@
 from lightgbm import LGBMRegressor
 from sklearn.model_selection import train_test_split
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.conformity_scores import (GammaConformityScore,
                                      ResidualNormalisedScore)
 from mapie.metrics import (hsic, regression_coverage_score_v2, regression_ssc,

examples/regression/2-advanced-analysis/plot_timeseries_enbpi.py

@@ -34,7 +34,7 @@
 from sklearn.ensemble import RandomForestRegressor
 from sklearn.model_selection import RandomizedSearchCV, TimeSeriesSplit
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.metrics import (regression_coverage_score,
                            regression_mean_width_score)
 from mapie.regression import MapieTimeSeriesRegressor

examples/regression/3-scientific-articles/plot_barber2020_simulations.py

@@ -35,7 +35,7 @@
 from matplotlib import pyplot as plt
 from sklearn.linear_model import LinearRegression
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.metrics import (regression_coverage_score,
                            regression_mean_width_score)
 from mapie_v1.regression import CrossConformalRegressor

examples/regression/3-scientific-articles/plot_kim2020_simulations.py

@@ -45,7 +45,7 @@
 from sklearn.linear_model import Ridge
 from sklearn.model_selection import train_test_split
 
-from mapie._typing import ArrayLike, NDArray
+from numpy.typing import ArrayLike, NDArray
 from mapie.metrics import (regression_coverage_score,
                            regression_mean_width_score)
 from mapie_v1.regression import (

examples/regression/3-scientific-articles/plot_zaffran2022_comparison.py

@@ -45,7 +45,7 @@
 from sklearn.ensemble import RandomForestRegressor
 from sklearn.model_selection import PredefinedSplit
 
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.conformity_scores import AbsoluteConformityScore
 from mapie.time_series_regression import MapieTimeSeriesRegressor
 

mapie/aggregation_functions.py

@@ -2,7 +2,7 @@
 
 import numpy as np
 
-from ._typing import NDArray
+from numpy.typing import NDArray
 
 
 def phi1D(

mapie/calibration.py

@@ -12,7 +12,7 @@
 from sklearn.utils.validation import (_check_y, _num_samples, check_is_fitted,
                                       indexable)
 
-from ._typing import ArrayLike, NDArray
+from numpy.typing import ArrayLike, NDArray
 from .utils import (check_estimator_classification,
                     check_estimator_fit_predict, check_n_features_in,
                     check_null_weight, fit_estimator, get_calib_set)

mapie/classification.py

@@ -10,7 +10,7 @@
 from sklearn.utils import check_random_state
 from sklearn.utils.validation import (_check_y, check_is_fitted, indexable)
 
-from mapie._typing import ArrayLike, NDArray
+from numpy.typing import ArrayLike, NDArray
 from mapie.conformity_scores import BaseClassificationScore
 from mapie.conformity_scores.sets.raps import RAPSConformityScore
 from mapie.conformity_scores.utils import (

mapie/conformity_scores/bounds/absolute.py

@@ -1,6 +1,6 @@
 import numpy as np
 
-from mapie._typing import ArrayLike, NDArray
+from numpy.typing import ArrayLike, NDArray
 from mapie.conformity_scores import BaseRegressionScore
 
 

mapie/conformity_scores/bounds/gamma.py

@@ -1,6 +1,6 @@
 import numpy as np
 
-from mapie._typing import ArrayLike, NDArray
+from numpy.typing import ArrayLike, NDArray
 from mapie.conformity_scores import BaseRegressionScore
 
 

mapie/conformity_scores/bounds/residuals.py

@@ -9,7 +9,7 @@
 from sklearn.utils.validation import (check_is_fitted, check_random_state,
                                       indexable)
 
-from mapie._typing import ArrayLike, NDArray
+from numpy.typing import ArrayLike, NDArray
 from mapie.conformity_scores import BaseRegressionScore
 
 

mapie/conformity_scores/classification.py

@@ -6,7 +6,7 @@
 from mapie.conformity_scores.interface import BaseConformityScore
 from mapie.estimator.classifier import EnsembleClassifier
 
-from mapie._typing import ArrayLike, NDArray
+from numpy.typing import ArrayLike, NDArray
 
 
 class BaseClassificationScore(BaseConformityScore, metaclass=ABCMeta):

mapie/conformity_scores/conformity_scores.py

@@ -4,7 +4,7 @@
 
 from mapie.conformity_scores.regression import BaseConformityScore
 from mapie._machine_precision import EPSILON
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 
 
 @deprecated(

mapie/conformity_scores/interface.py

@@ -4,8 +4,7 @@
 import numpy as np
 from sklearn.base import BaseEstimator
 
-from mapie._compatibility import np_nanquantile
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 
 
 class BaseConformityScore(metaclass=ABCMeta):
@@ -155,7 +154,7 @@ def get_quantile(
         # Otherwise, the quantile is calculated as the corrected lower quantile
         # of the signed conformity scores.
         quantile = signed * np.column_stack([
-            np_nanquantile(
+            np.nanquantile(
                 signed * conformity_scores, _alpha_cor,
                 axis=axis, method="lower"
             ) if not (unbounded and _alpha >= 1) else np.inf * np.ones(n_ref)

mapie/conformity_scores/regression.py

@@ -7,9 +7,8 @@
 from mapie.conformity_scores.interface import BaseConformityScore
 from mapie.estimator.regressor import EnsembleRegressor
 
-from mapie._compatibility import np_nanquantile
 from mapie._machine_precision import EPSILON
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 
 
 class BaseRegressionScore(BaseConformityScore, metaclass=ABCMeta):
@@ -238,13 +237,13 @@ def _beta_optimize(
                 num=len(lower_bounds),
                 endpoint=True,
             )
-            one_alpha_beta = np_nanquantile(
+            one_alpha_beta = np.nanquantile(
                 upper_bounds.astype(float),
                 1 - _alpha + betas,
                 axis=1,
                 method="higher",
             )
-            beta = np_nanquantile(
+            beta = np.nanquantile(
                 lower_bounds.astype(float),
                 betas,
                 axis=1,

mapie/conformity_scores/sets/aps.py

@@ -11,7 +11,7 @@
 from mapie.estimator.classifier import EnsembleClassifier
 
 from mapie._machine_precision import EPSILON
-from mapie._typing import ArrayLike, NDArray
+from numpy.typing import ArrayLike, NDArray
 from mapie.utils import compute_quantiles
 
 

mapie/conformity_scores/sets/lac.py

@@ -7,7 +7,7 @@
 from mapie.estimator.classifier import EnsembleClassifier
 
 from mapie._machine_precision import EPSILON
-from mapie._typing import NDArray
+from numpy.typing import NDArray
 from mapie.utils import compute_quantiles