Merge branch 'master' into 503-404-citations-file-in-readthedocs

Author: LacombeLouis

HISTORY.rst

@@ -6,6 +6,8 @@ History
 ------------------
 
 * Fix citations and license links
+* Fix the CQR tutorial to have same data in both methods
+* Add `** predict_params` in fit and predict method for Mapie Classifier
 * Add Mondrian Conformal Prediction for regression and classification
 * Add `** predict_params` in fit and predict method for Mapie Regression
 * Update the ts-changepoint notebook with the tutorial

examples/regression/4-tutorials/plot_cqr_tutorial.py

@@ -101,11 +101,6 @@ class :class:`~mapie.subsample.Subsample` (note that the `alpha` parameter is
     y['MedHouseVal'],
     random_state=random_state
 )
-X_train, X_calib, y_train, y_calib = train_test_split(
-    X_train,
-    y_train,
-    random_state=random_state
-)
 
 
 ##############################################################################
@@ -267,13 +262,19 @@ def plot_prediction_intervals(
     if strategy == "cqr":
         mapie = MapieQuantileRegressor(estimator, **params)
         mapie.fit(
-            X_train, y_train,
-            X_calib=X_calib, y_calib=y_calib,
+            X_train,
+            y_train,
+            calib_size=0.3,
             random_state=random_state
         )
         y_pred[strategy], y_pis[strategy] = mapie.predict(X_test)
     else:
-        mapie = MapieRegressor(estimator, **params, random_state=random_state)
+        mapie = MapieRegressor(
+            estimator,
+            test_size=0.3,
+            random_state=random_state,
+            **params
+        )
         mapie.fit(X_train, y_train)
         y_pred[strategy], y_pis[strategy] = mapie.predict(X_test, alpha=0.2)
     (

mapie/classification.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import warnings
-from typing import Iterable, Optional, Tuple, Union, cast
+from typing import Any, Iterable, Optional, Tuple, Union, cast
 
 import numpy as np
 from sklearn.base import BaseEstimator, ClassifierMixin
@@ -20,7 +20,8 @@
 from mapie.estimator.classifier import EnsembleClassifier
 from mapie.utils import (check_alpha, check_alpha_and_n_samples, check_cv,
                          check_estimator_classification, check_n_features_in,
-                         check_n_jobs, check_null_weight, check_verbose)
+                         check_n_jobs, check_null_weight, check_predict_params,
+                         check_verbose)
 
 
 class MapieClassifier(BaseEstimator, ClassifierMixin):
@@ -419,7 +420,7 @@ def fit(
         sample_weight: Optional[ArrayLike] = None,
         size_raps: Optional[float] = None,
         groups: Optional[ArrayLike] = None,
-        **fit_params,
+        **kwargs: Any
     ) -> MapieClassifier:
         """
         Fit the base estimator or use the fitted base estimator.
@@ -453,14 +454,22 @@ def fit(
 
             By default ``None``.
 
-        **fit_params : dict
-            Additional fit parameters.
+        kwargs : dict
+            Additional fit and predict parameters.
 
         Returns
         -------
         MapieClassifier
             The model itself.
         """
+        fit_params = kwargs.pop('fit_params', {})
+        predict_params = kwargs.pop('predict_params', {})
+
+        if len(predict_params) > 0:
+            self._predict_params = True
+        else:
+            self._predict_params = False
+
         # Checks
         (estimator,
          self.conformity_score_function_,
@@ -496,7 +505,7 @@ def fit(
 
         # Predict on calibration data
         y_pred_proba, y, y_enc = self.estimator_.predict_proba_calib(
-            X, y, y_enc, groups
+            X, y, y_enc, groups, **predict_params
         )
 
         # Compute the conformity scores
@@ -506,15 +515,15 @@ def fit(
                 y, y_pred_proba, y_enc=y_enc, X=X,
                 sample_weight=sample_weight, groups=groups
             )
-
         return self
 
     def predict(
         self,
         X: ArrayLike,
         alpha: Optional[Union[float, Iterable[float]]] = None,
         include_last_label: Optional[Union[bool, str]] = True,
-        agg_scores: Optional[str] = "mean"
+        agg_scores: Optional[str] = "mean",
+        **predict_params
     ) -> Union[NDArray, Tuple[NDArray, NDArray]]:
         """
         Prediction and prediction sets on new samples based on target
@@ -571,6 +580,9 @@ def predict(
 
             By default "mean".
 
+        predict_params : dict
+            Additional predict parameters.
+
         Returns
         -------
         Union[NDArray, Tuple[NDArray, NDArray]]
@@ -581,11 +593,16 @@ def predict(
         (n_samples,) and (n_samples, n_classes, n_alpha) if alpha is not None.
         """
         # Checks
+
+        if hasattr(self, '_predict_params'):
+            check_predict_params(self._predict_params,
+                                 predict_params, self.cv)
+
         check_is_fitted(self, self.fit_attributes)
         alpha = cast(Optional[NDArray], check_alpha(alpha))
 
         # Estimate predictions
-        y_pred = self.estimator_.single_estimator_.predict(X)
+        y_pred = self.estimator_.single_estimator_.predict(X, **predict_params)
         if alpha is None:
             return y_pred
 

mapie/estimator/classifier.py

@@ -222,6 +222,7 @@ def _predict_proba_oof_estimator(
         self,
         estimator: ClassifierMixin,
         X: ArrayLike,
+        **predict_params
     ) -> NDArray:
         """
         Predict probabilities of a test set from a fitted estimator.
@@ -239,7 +240,7 @@ def _predict_proba_oof_estimator(
         ArrayLike
             Predicted probabilities.
         """
-        y_pred_proba = estimator.predict_proba(X)
+        y_pred_proba = estimator.predict_proba(X, **predict_params)
         # we enforce y_pred_proba to contain all labels included in y
         if len(estimator.classes_) != self.n_classes:
             y_pred_proba = fix_number_of_classes(
@@ -252,7 +253,8 @@ def _predict_proba_calib_oof_estimator(
         estimator: ClassifierMixin,
         X: ArrayLike,
         val_index: ArrayLike,
-        k: int
+        k: int,
+        **predict_params
     ) -> Tuple[NDArray, ArrayLike, ArrayLike]:
         """
         Perform predictions on a single out-of-fold model on a validation set.
@@ -276,7 +278,8 @@ def _predict_proba_calib_oof_estimator(
 
         X_val = _safe_indexing(X, val_index)
         if _num_samples(X_val) > 0:
-            y_pred_proba = self._predict_proba_oof_estimator(estimator, X_val)
+            y_pred_proba = self._predict_proba_oof_estimator(estimator, X_val,
+                                                             **predict_params)
         else:
             y_pred_proba = np.array([])
         val_id = np.full(len(X_val), k, dtype=int)
@@ -401,6 +404,9 @@ def predict_proba_calib(
 
             By default ``None``.
 
+        **predict_params : dict
+            Additional predict parameters.
+
         Returns
         -------
         NDArray of shape (n_samples_test, 1)
@@ -417,7 +423,7 @@ def predict_proba_calib(
             cv = cast(BaseCrossValidator, self.cv)
             outputs = Parallel(n_jobs=self.n_jobs, verbose=self.verbose)(
                 delayed(self._predict_proba_calib_oof_estimator)(
-                    estimator, X, calib_index, k
+                    estimator, X, calib_index, k, **predict_params
                 )
                 for k, ((_, calib_index), estimator) in enumerate(
                     zip(cv.split(X, y, groups), self.estimators_)
@@ -462,6 +468,9 @@ def predict(
             How to aggregate the scores output by the estimators on test data
             if a cross-validation strategy is used
 
+        **predict_params : dict
+            Additional predict parameters.
+
         Returns
         -------
         NDArray
@@ -471,14 +480,15 @@ def predict(
         check_is_fitted(self, self.fit_attributes)
 
         if self.cv == "prefit":
-            y_pred_proba = self.single_estimator_.predict_proba(X)
+            y_pred_proba = self.single_estimator_.predict_proba(
+                X, **predict_params
+            )
         else:
             y_pred_proba_k = np.asarray(
-                Parallel(
-                    n_jobs=self.n_jobs, verbose=self.verbose
-                )(
-                    delayed(self._predict_proba_oof_estimator)(estimator, X)
-                    for estimator in self.estimators_
+                Parallel(n_jobs=self.n_jobs, verbose=self.verbose)(
+                    delayed(self._predict_proba_oof_estimator)(
+                        estimator, X, **predict_params
+                    ) for estimator in self.estimators_
                 )
             )
             if agg_scores == "crossval":