coverage 100%

Author: LacombeLouis

examples/regression/1-quickstart/plot_prefit.py

@@ -83,7 +83,7 @@ def f(x: NDArray) -> NDArray:
 estimator = LGBMRegressor(
     objective='quantile',
     alpha=0.5,
-) # Note that this is the same model as used for QR
+)  # Note that this is the same model as used for QR
 estimator.fit(X_train.reshape(-1, 1), y_train)
 list_estimators.append(estimator)
 

mapie/quantile_regression.py

@@ -22,6 +22,7 @@
     fit_estimator,
     check_lower_upper_bounds,
     check_defined_variables_predict_cqr,
+    check_estimator_fit_predict,
 )
 from ._compatibility import np_quantile
 from .regression import MapieRegressor
@@ -190,6 +191,12 @@ def _check_alpha(
         ValueError
             If the value of alpha is not between 0 and 1.0.
         """
+        if self.cv == "prefit":
+            warnings.warn(
+                "WARNING: The alpha that is set needs to be the same"
+                + " as the alpha of your prefitted model in the following"
+                " order [alpha/2, 1 - alpha/2, 0.5]"
+            )
         if isinstance(alpha, float):
             if np.any(np.logical_or(alpha <= 0, alpha >= 1.0)):
                 raise ValueError(
@@ -251,11 +258,7 @@ def _check_estimator(
                 solver="highs-ds",
                 alpha=0.0,
             )
-        if not (hasattr(estimator, "fit") and hasattr(estimator, "predict")):
-            raise ValueError(
-                "Invalid estimator. "
-                "Please provide a regressor with fit and predict methods."
-            )
+        check_estimator_fit_predict(estimator)
         if isinstance(estimator, Pipeline):
             self._check_estimator(estimator[-1])
             return estimator
@@ -408,10 +411,6 @@ def _check_calib_set(
     def _check_prefit_params(
         self,
         estimator: List[Union[RegressorMixin, Pipeline]],
-        X: ArrayLike,
-        y: ArrayLike,
-        X_calib: Optional[ArrayLike] = None,
-        y_calib: Optional[ArrayLike] = None,
     ) -> None:
         """
         Check the parameters set for the specific case of prefit
@@ -451,19 +450,14 @@ def _check_prefit_params(
             )
         if len(estimator) == 3:
             for est in estimator:
-                self._check_estimator(est)
+                check_estimator_fit_predict(est)
+                check_is_fitted(est)
         else:
             raise ValueError(
                     "You need to have provided 3 different estimators, they"
                     " need to be preset with alpha values in the following"
                     " order [alpha/2, 1 - alpha/2, 0.5]."
                     )
-        if self.alpha is not None:
-            warnings.warn(
-                "WARNING: The alpha that is set needs to be the same"
-                + " as the alpha of your prefitted model in the following"
-                " order [alpha/2, 1 - alpha/2, 0.5]"
-            )
 
     def fit(
         self,
@@ -536,12 +530,11 @@ def fit(
         self.estimators_: List[RegressorMixin] = []
         if self.cv == "prefit":
             estimator = cast(List, self.estimator)
-            self._check_prefit_params(estimator, X, y)
+            alpha = self._check_alpha(self.alpha)
+            self._check_prefit_params(estimator)
             X_calib, y_calib = indexable(X, y)
 
             self.n_calib_samples = _num_samples(y_calib)
-            check_alpha_and_n_samples(self.alpha, self.n_calib_samples)
-
             y_calib_preds = np.full(
                 shape=(3, self.n_calib_samples),
                 fill_value=np.nan

mapie/tests/test_quantile_regression.py

@@ -519,6 +519,120 @@ def test_linear_regression_results(strategy: str) -> None:
     np.testing.assert_allclose(coverage, COVERAGES[strategy], rtol=1e-2)
 
 
+def test_quantile_prefit_non_list() -> None:
+    """
+    Test that there is a list of estimators provided when cv='prefit'
+    is called for MapieQuantileRegressor.
+    """
+    with pytest.raises(
+        ValueError,
+        match=r".*Estimator for prefit must be an iterable object.*",
+    ):
+        not_an_iterable = 10
+        mapie_reg = MapieQuantileRegressor(
+            estimator=not_an_iterable,
+            cv="prefit"
+        )
+        mapie_reg.fit(
+            X_calib_toy,
+            y_calib_toy
+        )
+
+
+def test_quantile_prefit_three_estimators() -> None:
+    """
+    Test that there is a list of estimators three estimators provided for
+    cv="prefit".
+    """
+    with pytest.raises(
+        ValueError,
+        match=r".*You need to have provided 3 different estimators, th*",
+    ):
+        gb_trained1, gb_trained2 = clone(gb), clone(gb)
+        gb_trained1.fit(X_train, y_train)
+        gb_trained2.fit(X_train, y_train)
+        list_estimators = [gb_trained1, gb_trained2]
+        mapie_reg = MapieQuantileRegressor(
+            estimator=list_estimators,
+            cv="prefit"
+        )
+        mapie_reg.fit(
+            X_calib,
+            y_calib
+        )
+
+
+def test_prefit_no_fit_predict() -> None:
+    """
+    Check that the user is warned that the alphas need to be correctly set.
+    """
+    with pytest.raises(
+        ValueError,
+        match=r"Invalid estimator. Please provide a regressor with fit and*",
+    ):
+        gb_trained1, gb_trained2 = clone(gb), clone(gb)
+        gb_trained1.fit(X_train, y_train)
+        gb_trained2.fit(X_train, y_train)
+        gb_trained3 = 3
+        list_estimators = [gb_trained1, gb_trained2, gb_trained3]
+        mapie_reg = MapieQuantileRegressor(
+            estimator=list_estimators,
+            cv="prefit",
+            alpha=0.3
+        )
+        mapie_reg.fit(
+            X_calib,
+            y_calib
+        )
+
+
+def test_non_trained_estimator() -> None:
+    """
+    Check that the user is warned that the alphas need to be correctly set.
+    """
+    with pytest.raises(
+        ValueError,
+        match=r".*instance is not fitted yet. Call 'fit' with appropriate*",
+    ):
+        gb_trained1, gb_trained2, gb_trained3 = clone(gb), clone(gb), clone(gb)
+        gb_trained1.fit(X_train, y_train)
+        gb_trained2.fit(X_train, y_train)
+        list_estimators = [gb_trained1, gb_trained2, gb_trained3]
+        mapie_reg = MapieQuantileRegressor(
+            estimator=list_estimators,
+            cv="prefit",
+            alpha=0.3
+        )
+        mapie_reg.fit(
+            X_calib,
+            y_calib
+        )
+
+
+def test_warning_alpha_prefit() -> None:
+    """
+    Check that the user is warned that the alphas need to be correctly set.
+    """
+    with pytest.warns(
+        UserWarning,
+        match=r".*WARNING: The alpha that is set needs to be the same*"
+    ):
+        gb_trained1, gb_trained2, gb_trained3 = clone(gb), clone(gb), clone(gb)
+        gb_trained1.fit(X_train, y_train)
+        gb_trained2.fit(X_train, y_train)
+        gb_trained3.fit(X_train, y_train)
+        list_estimators = [gb_trained1, gb_trained2, gb_trained3]
+        mapie_reg = MapieQuantileRegressor(
+            estimator=list_estimators,
+            cv="prefit",
+            alpha=0.3
+        )
+        mapie_reg.fit(
+            X_calib,
+            y_calib
+        )
+
+
 @pytest.mark.parametrize("estimator", ESTIMATOR)
 def test_pipeline_compatibility(estimator: RegressorMixin) -> None:
     """Check that MAPIE works on pipeline based on pandas dataframes"""

mapie/utils.py

@@ -563,3 +563,26 @@ def check_defined_variables_predict_cqr(
             "WARNING: Alpha should not be specified in the prediction method\n"
             + "with conformalized quantile regression."
         )
+
+
+def check_estimator_fit_predict(
+    estimator: Union[RegressorMixin, ClassifierMixin]
+) -> None:
+    """
+    Check that the estimator has a fit and precict method.
+
+    Parameters
+    ----------
+    estimator : Union[RegressorMixin, ClassifierMixin]
+        Estimator to train.
+
+    Raises
+    ------
+    ValueError
+        If the estimator does not have a fit or predict attribute.
+    """
+    if not (hasattr(estimator, "fit") and hasattr(estimator, "predict")):
+        raise ValueError(
+            "Invalid estimator. "
+            "Please provide a regressor with fit and predict methods."
+        )