Merge branch 'main' into refactor-scenarios-lazy

.all-contributorsrc

@@ -2116,7 +2116,8 @@
       "profile": "https://www.linkedin.com/in/pranav-prajapati-a5b413226/",
       "contributions": [
         "code",
-        "test"
+        "test",
+        "bug"
       ]
     },
     {
@@ -2342,7 +2343,9 @@
       "profile": "https://github.com/Ram0nB",
       "contributions": [
         "doc",
-        "code"
+        "code",
+        "bug",
+        "test"
       ]
     },
     {
@@ -2735,6 +2738,15 @@
         "doc"
       ]
     },
+    {
+      "login": "morestart",
+      "name": "ctl",
+      "avatar_url": "https://avatars.githubusercontent.com/u/35556811",
+      "profile": "https://github.com/morestart",
+      "contributions": [
+        "bug"
+      ]
+    },
     {
       "login": "anteemony",
       "name": "Anthony Okonneh",

extension_templates/alignment.py

@@ -82,7 +82,10 @@ class MyAligner(BaseAligner):
         "authors": ["author1", "author2"],  # authors, GitHub handles
         "maintainers": ["maintainer1", "maintainer2"],  # maintainers, GitHub handles
         # author = significant contribution to code at some point
-        # maintainer = algorithm maintainer role, "owner"
+        #     if interfacing a 3rd party estimator, ensure to give credit to the
+        #     authors of the interfaced estimator
+        # maintainer = algorithm maintainer role, "owner" of the sktime class
+        #     for 3rd party interfaces, the scope is the sktime class only
         # specify one or multiple authors and maintainers, only for sktime contribution
         # remove maintainer tag if maintained by sktime core team
         #

extension_templates/classification.py

@@ -82,8 +82,11 @@ class MyTimeSeriesClassifier(BaseClassifier):
         "authors": ["author1", "author2"],  # authors, GitHub handles
         "maintainers": ["maintainer1", "maintainer2"],  # maintainers, GitHub handles
         # author = significant contribution to code at some point
-        # maintainer = algorithm maintainer role, "owner"
-        # specify one or multiple authors and maintainers, only for sktime contribution
+        #     if interfacing a 3rd party estimator, ensure to give credit to the
+        #     authors of the interfaced estimator
+        # maintainer = algorithm maintainer role, "owner" of the sktime class
+        #     for 3rd party interfaces, the scope is the sktime class only
+        # specify one or multiple authors and maintainers
         # remove maintainer tag if maintained by sktime core team
         #
         "python_version": None,  # PEP 440 python version specifier to limit versions

extension_templates/clustering.py

@@ -77,7 +77,10 @@ class MyClusterer(BaseClusterer):
         "authors": ["author1", "author2"],  # authors, GitHub handles
         "maintainers": ["maintainer1", "maintainer2"],  # maintainers, GitHub handles
         # author = significant contribution to code at some point
-        # maintainer = algorithm maintainer role, "owner"
+        #     if interfacing a 3rd party estimator, ensure to give credit to the
+        #     authors of the interfaced estimator
+        # maintainer = algorithm maintainer role, "owner" of the sktime class
+        #     for 3rd party interfaces, the scope is the sktime class only
         # specify one or multiple authors and maintainers, only for sktime contribution
         # remove maintainer tag if maintained by sktime core team
         #

extension_templates/dist_kern_panel.py

@@ -62,7 +62,10 @@ class MyTrafoPwPanel(BasePairwiseTransformerPanel):
         "authors": ["author1", "author2"],  # authors, GitHub handles
         "maintainers": ["maintainer1", "maintainer2"],  # maintainers, GitHub handles
         # author = significant contribution to code at some point
-        # maintainer = algorithm maintainer role, "owner"
+        #     if interfacing a 3rd party estimator, ensure to give credit to the
+        #     authors of the interfaced estimator
+        # maintainer = algorithm maintainer role, "owner" of the sktime class
+        #     for 3rd party interfaces, the scope is the sktime class only
         # remove maintainer tag if maintained by sktime core team
     }
     # in case of inheritance, concrete class should typically set tags

extension_templates/dist_kern_tab.py

@@ -62,7 +62,10 @@ class MyTrafoPw(BasePairwiseTransformer):
         "authors": ["author1", "author2"],  # authors, GitHub handles
         "maintainers": ["maintainer1", "maintainer2"],  # maintainers, GitHub handles
         # author = significant contribution to code at some point
-        # maintainer = algorithm maintainer role, "owner"
+        #     if interfacing a 3rd party estimator, ensure to give credit to the
+        #     authors of the interfaced estimator
+        # maintainer = algorithm maintainer role, "owner" of the sktime class
+        #     for 3rd party interfaces, the scope is the sktime class only
         # remove maintainer tag if maintained by sktime core team
     }
     # in case of inheritance, concrete class should typically set tags

extension_templates/forecasting.py

@@ -165,10 +165,13 @@ class MyForecaster(BaseForecaster):
         # valid values: str or list of str, should be GitHub handles
         # this should follow best scientific contribution practices
         # scope is the code, not the methodology (method is per paper citation)
+        # if interfacing a 3rd party estimator, ensure to give credit to the
+        # authors of the interfaced estimator
         #
         # maintainer = current maintainer(s) of the estimator
         # per algorithm maintainer role, see governance document
         # this is an "owner" type role, with rights and maintenance duties
+        # for 3rd party interfaces, the scope is the sktime class only
         "maintainers": ["maintainer1", "maintainer2"],
         # valid values: str or list of str, should be GitHub handles
         # remove tag if maintained by sktime core team

extension_templates/forecasting_simple.py

@@ -117,10 +117,13 @@ class MyForecaster(BaseForecaster):
         # valid values: str or list of str, should be GitHub handles
         # this should follow best scientific contribution practices
         # scope is the code, not the methodology (method is per paper citation)
+        # if interfacing a 3rd party estimator, ensure to give credit to the
+        # authors of the interfaced estimator
         #
         # maintainer = current maintainer(s) of the estimator
         # per algorithm maintainer role, see governance document
         # this is an "owner" type role, with rights and maintenance duties
+        # for 3rd party interfaces, the scope is the sktime class only
         "maintainers": ["maintainer1", "maintainer2"],
         # valid values: str or list of str, should be GitHub handles
         # remove tag if maintained by sktime core team

extension_templates/forecasting_supersimple.py

@@ -76,7 +76,10 @@ class MyForecaster(BaseForecaster):
         "authors": ["author1", "author2"],  # authors, GitHub handles
         "maintainers": ["maintainer1", "maintainer2"],  # maintainers, GitHub handles
         # author = significant contribution to code at some point
-        # maintainer = algorithm maintainer role, "owner"
+        #     if interfacing a 3rd party estimator, ensure to give credit to the
+        #     authors of the interfaced estimator
+        # maintainer = algorithm maintainer role, "owner" of the sktime class
+        #     for 3rd party interfaces, the scope is the sktime class only
         # remove maintainer tag if maintained by sktime core team
         #
         # do not change these:

extension_templates/split.py

@@ -101,10 +101,13 @@ class MySplitter(BaseSplitter):
         # valid values: str or list of str, should be GitHub handles
         # this should follow best scientific contribution practices
         # scope is the code, not the methodology (method is per paper citation)
+        # if interfacing a 3rd party estimator, ensure to give credit to the
+        # authors of the interfaced estimator
         #
         # maintainer = current maintainer(s) of the estimator
         # per algorithm maintainer role, see governance document
         # this is an "owner" type role, with rights and maintenance duties
+        # for 3rd party interfaces, the scope is the sktime class only
         "maintainers": ["maintainer1", "maintainer2"],
         # valid values: str or list of str, should be GitHub handles
         # remove tag if maintained by sktime core team

extension_templates/transformer.py

@@ -257,10 +257,13 @@ class MyTransformer(BaseTransformer):
         # valid values: str or list of str, should be GitHub handles
         # this should follow best scientific contribution practices
         # scope is the code, not the methodology (method is per paper citation)
+        # if interfacing a 3rd party estimator, ensure to give credit to the
+        # authors of the interfaced estimator
         #
         # maintainer = current maintainer(s) of the estimator
         # per algorithm maintainer role, see governance document
         # this is an "owner" type role, with rights and maintenance duties
+        # for 3rd party interfaces, the scope is the sktime class only
         "maintainers": ["maintainer1", "maintainer2"],
         # valid values: str or list of str, should be GitHub handles
         # remove tag if maintained by sktime core team

extension_templates/transformer_simple.py

@@ -182,10 +182,13 @@ class MyTransformer(BaseTransformer):
         # valid values: str or list of str, should be GitHub handles
         # this should follow best scientific contribution practices
         # scope is the code, not the methodology (method is per paper citation)
+        # if interfacing a 3rd party estimator, ensure to give credit to the
+        # authors of the interfaced estimator
         #
         # maintainer = current maintainer(s) of the estimator
         # per algorithm maintainer role, see governance document
         # this is an "owner" type role, with rights and maintenance duties
+        # for 3rd party interfaces, the scope is the sktime class only
         "maintainers": ["maintainer1", "maintainer2"],
         # valid values: str or list of str, should be GitHub handles
         # remove tag if maintained by sktime core team

extension_templates/transformer_supersimple.py

@@ -79,7 +79,10 @@ class MyTransformer(BaseTransformer):
         "authors": ["author1", "author2"],  # authors, GitHub handles
         "maintainers": ["maintainer1", "maintainer2"],  # maintainers, GitHub handles
         # author = significant contribution to code at some point
-        # maintainer = algorithm maintainer role, "owner"
+        #     if interfacing a 3rd party estimator, ensure to give credit to the
+        #     authors of the interfaced estimator
+        # maintainer = algorithm maintainer role, "owner" of the sktime class
+        #     for 3rd party interfaces, the scope is the sktime class only
         # remove maintainer tag if maintained by sktime core team
         #
         # do not change these:

pyproject.toml

@@ -80,7 +80,7 @@ all_extras = [
   "arch>=5.6,<6.4.0",
   "cloudpickle",
   "dash!=2.9.0",
-  "dask<2024.2.2",
+  "dask<2024.4.2; extra == 'dataframe'",
   "dtw-python",
   'esig==0.9.7; python_version < "3.10"',
   'filterpy>=1.4.5; python_version < "3.11"',
@@ -121,7 +121,7 @@ all_extras_pandas2 = [
   "arch>=5.6,<6.4.0",
   "cloudpickle",
   "dash!=2.9.0",
-  "dask<2024.2.2",
+  "dask<2024.4.2; extra == 'dataframe'",
   "dtw-python",
   'esig==0.9.7; python_version < "3.10"',
   'filterpy>=1.4.5; python_version < "3.11"',

sktime/datasets/_single_problem_loaders.py

@@ -1413,10 +1413,7 @@ def load_forecastingdata(
         # valid dataset names for classification, regression, forecasting datasets repo
         if name not in list(tsf_all_datasets):
             raise ValueError(
-                {name}
-                + " is not a valid dataset name. \
-                    List of valid dataset names can be found at \
-                    sktime.datasets.tsf_dataset_names.tsf_all_datasets"
+                f"Error in load_forecastingdata, Invalid dataset name = {name}."
             )
 
         url = f"https://zenodo.org/record/{tsf_all[name]}/files/{name}.zip"

sktime/datasets/tests/test_datadownload.py

@@ -80,3 +80,14 @@ def test_check_link_downloadable(name):
 
     assert "application/octet-stream" in content_type, "URL is not downloadable."
     assert "attachment" in content_disposition, "URL is not downloadable."
+
+
+@pytest.mark.datadownload
+@pytest.mark.parametrize("name", ["invalid_name"])
+def test_load_forecasting_data_invalid_name(name):
+    """Test load_forecastingdata with invalid name."""
+    with pytest.raises(
+        ValueError,
+        match=f"Error in load_forecastingdata, Invalid dataset name = {name}.",
+    ):
+        load_forecastingdata(name=name)

sktime/regression/deep_learning/cnn.py

@@ -186,7 +186,7 @@ def _fit(self, X, y):
         self.input_shape = X.shape[1:]
         self.model_ = self.build_model(self.input_shape)
         if self.verbose:
-            self.model.summary()
+            self.model_.summary()
 
         self.history = self.model_.fit(
             X,
@@ -234,6 +234,7 @@ def get_test_params(cls, parameter_set="default"):
             "batch_size": 6,
             "kernel_size": 2,
             "n_conv_layers": 1,
+            "verbose": True,
         }
         test_params = [param1, param2]
 

sktime/transformations/series/impute.py

@@ -23,7 +23,9 @@ class Imputer(BaseTransformer):
     Parameters
     ----------
     method : str, default="drift"
-        Method to fill the missing values.
+        Method to fill the missing values. Not all methods can extrapolate, so after
+        ``method`` is applied the remaining missing values are filled with ``ffill``
+        then ``bfill``.
 
         * "drift" : drift/trend values by sktime.PolynomialTrendForecaster(degree=1)
           first, X in transform() is filled with ffill then bfill
@@ -231,22 +233,26 @@ def _transform(self, X, y=None):
         elif self.method == "constant":
             return X.fillna(value=self.value)
         elif isinstance(index, pd.MultiIndex):
-            X_grouped = X.groupby(level=list(range(index.nlevels - 1)))
+            X_group_levels = list(range(index.nlevels - 1))
 
             if self.method in ["backfill", "bfill"]:
-                X = X_grouped.bfill()
-                # fill trailing NAs of panel instances with reverse method
-                return X.ffill()
+                X = X.groupby(level=X_group_levels).bfill()
             elif self.method in ["pad", "ffill"]:
-                X = X_grouped.ffill()
-                # fill leading NAs of panel instances with reverse method
-                return X.bfill()
+                X = X.groupby(level=X_group_levels).ffill()
             elif self.method == "mean":
-                return X_grouped.fillna(value=self._mean)
+                X = X.groupby(level=X_group_levels).fillna(value=self._mean)
             elif self.method == "median":
-                return X_grouped.fillna(value=self._median)
+                X = X.groupby(level=X_group_levels).fillna(value=self._median)
             else:
                 raise AssertionError("Code should not be reached")
+
+            # fill first/last elements of series,
+            # as some methods can't impute those
+            X = X.groupby(level=X_group_levels).ffill()
+            X = X.groupby(level=X_group_levels).bfill()
+
+            return X
+
         else:
             if self.method in ["backfill", "bfill"]:
                 X = X.bfill()

sktime/transformations/series/tests/test_imputer.py

@@ -8,7 +8,9 @@
 import numpy as np
 import pytest
 
+from sktime.datatypes import get_examples
 from sktime.forecasting.naive import NaiveForecaster
+from sktime.transformations.compose import TransformByLevel
 from sktime.transformations.series.impute import Imputer
 from sktime.utils._testing.forecasting import make_forecasting_problem
 from sktime.utils._testing.hierarchical import _make_hierarchical
@@ -58,6 +60,44 @@ def test_imputer(method, Z, value, forecaster):
     assert not y_hat.isnull().to_numpy().any()
 
 
+@pytest.mark.parametrize(
+    "method",
+    [
+        "linear",
+        "nearest",
+        "mean",
+        "median",
+        "backfill",
+        "pad",
+    ],
+)
+def test_impute_multiindex(method):
+    """Test for data leakage in case of pd-multiindex data.
+
+    Failure case in bug #6224
+    """
+
+    df = get_examples(mtype="pd-multiindex")[0].copy()
+    df.iloc[:3, :] = np.nan  # instance 0 entirely missing
+    df.iloc[3:4, :] = np.nan  # instance 1 first timepoint missing
+    df.iloc[8:, :] = np.nan  # instance 2 last timepoint missing
+
+    imp = Imputer(method=method)
+    df_imp = imp.fit_transform(df)
+
+    # instance 0 entirely missing, so it should remain missing
+    assert np.array_equal(df.iloc[:3, :], df_imp.iloc[:3, :], equal_nan=True)
+
+    # instance 1 and 2 should not have any missing values
+    assert not df_imp.iloc[3:, :].isna().any().any()
+
+    # test consistency between applying the imputer to every instance separately,
+    # vs applying them to the panel
+    imp_tbl = TransformByLevel(Imputer(method=method))
+    df_imp_tbl = imp_tbl.fit_transform(df)
+    assert np.array_equal(df_imp, df_imp_tbl, equal_nan=True)
+
+
 def test_imputer_forecaster_y():
     """Test that forecaster imputer works with y.