refactor : rename partial_fit as internal method of TimeSeriesRegressor and update tests and docs (#792)

* rename partial_fit as internal method of TimeSeriesRegressor 

* update tests and doc

* add author

Author: hmaissoro

AUTHORS.rst

@@ -8,6 +8,7 @@ Development Lead
 * Vincent Blot <[email protected]>
 * Valentin Laurent <[email protected]>
 * Adrien Le Coz <[email protected]>
+* Hassan Maissoro <[email protected]>
 * Geoffray Brelurut <[email protected]>
 
 Emeritus Core Developers

examples/regression/1-quickstart/plot_ts-tutorial.py

@@ -43,7 +43,7 @@ class that block bootstraps the training set.
 each time new observations are available so that the deterioration of
 predictions, or the increase of noise level, can be dynamically taken into
 account. It can be done with :class:`~mapie.regression.TimeSeriesRegressor`
-through the ``partial_fit`` class method called at every step.
+through the ``update`` class method called at every step.
 
 
 The ACI strategy allows you to adapt the conformal inference
@@ -64,9 +64,9 @@ class that block bootstraps the training set.
 from sklearn.model_selection import RandomizedSearchCV, TimeSeriesSplit
 
 from mapie.metrics.regression import (
+    coverage_width_based,
     regression_coverage_score,
     regression_mean_width_score,
-    coverage_width_based,
 )
 from mapie.regression import TimeSeriesRegressor
 from mapie.subsample import BlockBootstrap
@@ -178,11 +178,11 @@ class that block bootstraps the training set.
 # - with a regular ``.fit`` and ``.predict`` process, limiting the use of
 #   trainining set residuals to build prediction intervals
 #
-# - using ``.partial_fit`` in addition to ``.fit`` and ``.predict`` allowing
+# - using ``.update`` in addition to ``.fit`` and ``.predict`` allowing
 #   MAPIE to use new residuals from the test points as new data are becoming
 #   available.
 #
-# - using ``.partial_fit`` and ``.adapt_conformal_inference`` in addition to
+# - using ``.update`` and ``.adapt_conformal_inference`` in addition to
 #   ``.fit`` and ``.predict`` allowing MAPIE to use new residuals from the
 #   test points as new data are becoming available.
 #
@@ -209,7 +209,7 @@ class that block bootstraps the training set.
 
 
 ##############################################################################
-# Let's start by estimating prediction intervals without partial fit.
+# Let's start by estimating prediction intervals without partial update of the residuals.
 
 # For EnbPI
 mapie_enbpi = mapie_enbpi.fit(X_train, y_train)
@@ -270,7 +270,7 @@ class that block bootstraps the training set.
 
 
 ##############################################################################
-# Let's now estimate prediction intervals with partial fit. As discussed
+# Let's now estimate prediction intervals with updated residuals. As discussed
 # previously, the update of the residuals and the one-step ahead predictions
 # are performed sequentially in a loop.
 
@@ -289,7 +289,7 @@ class that block bootstraps the training set.
 )
 
 for step in range(gap, len(X_test), gap):
-    mapie_enbpi.partial_fit(
+    mapie_enbpi.update(
         X_test.iloc[(step - gap) : step, :],
         y_test.iloc[(step - gap) : step],
     )
@@ -336,9 +336,9 @@ class that block bootstraps the training set.
 )
 
 for step in range(gap, len(X_test), gap):
-    mapie_aci.partial_fit(
-        X_test.iloc[(step - gap) : step, :],
-        y_test.iloc[(step - gap) : step],
+    mapie_aci.update(
+        X_test.iloc[(step - gap) : step, :].to_numpy(),
+        y_test.iloc[(step - gap) : step].to_numpy(),
     )
     mapie_aci.adapt_conformal_inference(
         X_test.iloc[(step - gap) : step, :].to_numpy(),
@@ -514,3 +514,5 @@ class that block bootstraps the training set.
 # One can notice that the uncertainty's explosion happens about one day late.
 # This is because enough new residuals are needed to change the quantiles
 # obtained from the residuals distribution.
+
+# %%

mapie/regression/time_series_regression.py

@@ -1,18 +1,20 @@
 from __future__ import annotations
 
-import warnings
 from typing import Iterable, Optional, Tuple, Union, cast
 
 import numpy as np
+from numpy.typing import ArrayLike, NDArray
 from sklearn.base import RegressorMixin
 from sklearn.model_selection import BaseCrossValidator
 from sklearn.utils.validation import check_is_fitted
 
-from numpy.typing import ArrayLike, NDArray
 from mapie.conformity_scores import BaseRegressionScore
 from mapie.regression.regression import _MapieRegressor
-from mapie.utils import _check_alpha, _check_gamma
-from mapie.utils import _transform_confidence_level_to_alpha_list
+from mapie.utils import (
+    _check_alpha,
+    _check_gamma,
+    _transform_confidence_level_to_alpha_list,
+)
 
 
 class TimeSeriesRegressor(_MapieRegressor):
@@ -24,7 +26,7 @@ class TimeSeriesRegressor(_MapieRegressor):
     Both strategies are estimating prediction intervals
     on single-output time series.
 
-    EnbPI allows you to update conformal scores using the ``partial_fit``
+    EnbPI allows you to update conformal scores using the ``update``
     function. It will replace the oldest one with the newest scores.
     It will keep the same amount of total scores
 
@@ -118,7 +120,7 @@ def _relative_conformity_scores(
         )
         return scores
 
-    def partial_fit(
+    def _update_conformity_scores_with_ensemble(
         self,
         X: ArrayLike,
         y: ArrayLike,
@@ -127,7 +129,7 @@ def partial_fit(
         """
         Update the ``conformity_scores_`` attribute when new data with known
         labels are available.
-        Note: Don't use ``partial_fit`` with samples of the training set.
+        Note: Don't use ``_update_conformity_scores_with_ensemble`` with samples of the training set.
 
         Parameters
         ----------
@@ -159,13 +161,6 @@ def partial_fit(
             If the length of ``y`` is greater than
             the length of the training set.
         """
-        warnings.warn(
-            "WARNING: Deprecated method. "
-            + 'The method "partial_fit" will be removed in v1.2. '
-            + 'Use "update" instead to keep '
-            + "the same behavior in the future.",
-            DeprecationWarning,
-        )
         check_is_fitted(self, self.fit_attributes)
         X, y = cast(NDArray, X), cast(NDArray, y)
         m, n = len(X), len(self.conformity_scores_)
@@ -330,8 +325,8 @@ def update(
     ) -> TimeSeriesRegressor:
         """
         Update with respect to the used ``method``.
-        ``method="enbpi"`` will call ``partial_fit`` method and
-        ``method="aci"`` will call ``adapt_conformal_inference`` method.
+        ``method="enbpi"`` updates conformity scores via EnbPI,
+        ``method="aci"`` calls ``adapt_conformal_inference``.
 
         Parameters
         ----------
@@ -381,7 +376,7 @@ def update(
         """
         self._check_method(self.method)
         if self.method == "enbpi":
-            return self.partial_fit(X, y, ensemble=ensemble)
+            return self._update_conformity_scores_with_ensemble(X, y, ensemble=ensemble)
         elif self.method == "aci":
             return self.adapt_conformal_inference(
                 X,

mapie/tests/test_time_series_regression.py

@@ -4,19 +4,17 @@
 
 import numpy as np
 import pytest
+from numpy.typing import NDArray
 from sklearn import __version__ as sklearn_version
 from sklearn.datasets import make_regression
 from sklearn.linear_model import LinearRegression
 from sklearn.model_selection import KFold, LeaveOneOut, train_test_split
 from sklearn.utils.estimator_checks import check_estimator
 from typing_extensions import TypedDict
 
-from numpy.typing import NDArray
 from mapie.aggregation_functions import aggregate_all
 from mapie.conformity_scores import AbsoluteConformityScore
-from mapie.metrics.regression import (
-    regression_coverage_score,
-)
+from mapie.metrics.regression import regression_coverage_score
 from mapie.regression import TimeSeriesRegressor
 from mapie.subsample import BlockBootstrap
 
@@ -357,28 +355,30 @@ def test_MapieTimeSeriesRegressor_if_alpha_is_None() -> None:
         y_pred, y_pis = mapie_ts_reg.predict(X_toy, confidence_level=None)
 
 
-def test_MapieTimeSeriesRegressor_partial_fit_ensemble() -> None:
-    """Test ``partial_fit``."""
+def test_MapieTimeSeriesRegressor_update_conformity_scores_with_ensemble() -> None:
+    """Test ``_update_conformity_scores_with_ensemble``"""
     mapie_ts_reg = TimeSeriesRegressor(method="enbpi", cv=-1)
     mapie_ts_reg.fit(X_toy, y_toy)
-    mapie_ts_reg.partial_fit(X_toy, y_toy, ensemble=True)
+    mapie_ts_reg._update_conformity_scores_with_ensemble(X_toy, y_toy, ensemble=True)
     assert round(mapie_ts_reg.conformity_scores_[-1], 2) == round(
         np.abs(CONFORMITY_SCORES[0]), 2
     )
-    mapie_ts_reg.partial_fit(
+    mapie_ts_reg._update_conformity_scores_with_ensemble(
         X=np.array([UPDATE_DATA[0]]), y=np.array([UPDATE_DATA[1]]), ensemble=True
     )
     assert round(mapie_ts_reg.conformity_scores_[-1], 2) == round(
         CONFORMITY_SCORES[1], 2
     )
 
 
-def test_MapieTimeSeriesRegressor_partial_fit_too_big() -> None:
-    """Test ``partial_fit`` raised error."""
+def test_MapieTimeSeriesRegressor_update_conformity_scores_with_ensemble_too_big() -> (
+    None
+):
+    """Test that ``_update_conformity_scores_with_ensemble`` raises an error."""
     mapie_ts_reg = TimeSeriesRegressor(method="enbpi", cv=-1)
     mapie_ts_reg.fit(X_toy, y_toy)
     with pytest.raises(ValueError, match=r".*The number of observations*"):
-        mapie_ts_reg = mapie_ts_reg.partial_fit(X=X, y=y)
+        mapie_ts_reg = mapie_ts_reg._update_conformity_scores_with_ensemble(X=X, y=y)
 
 
 def test_MapieTimeSeriesRegressor_beta_optimize_error() -> None:
@@ -458,14 +458,6 @@ def test_aci__get_alpha_with_unknown_alpha() -> None:
     np.testing.assert_allclose(mapie_ts_reg.current_alpha[0.2], 0.3, rtol=1e-3)
 
 
-def test_deprecated_partial_fit_warning() -> None:
-    """Test that a warning is raised if use partial_fit"""
-    mapie_ts_reg = TimeSeriesRegressor(method="enbpi", cv=-1)
-    mapie_ts_reg.fit(X_toy, y_toy)
-    with pytest.warns(DeprecationWarning, match=r".*WARNING: Deprecated method.*"):
-        mapie_ts_reg = mapie_ts_reg.partial_fit(X_toy, y_toy)
-
-
 @pytest.mark.parametrize("method", ["wrong_method"])
 def test_method_error_in_update(monkeypatch: Any, method: str) -> None:
     """Test else condition for the method in .update"""