Merge branch 'main' of github.com:decarsg/pykoop into paper

Author: sdahdah

.github/CODEOWNERS

@@ -0,0 +1 @@
+* @sdahdah

.github/workflows/test-package-no-mosek.yml

@@ -10,7 +10,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ['3.8', '3.9', '3.10', '3.11']
+        python-version: ['3.8', '3.9', '3.10', '3.11', '3.12']
     steps:
     - uses: actions/checkout@v2
     - name: Set up Python ${{matrix.python-version}}

.github/workflows/test-package.yml

@@ -10,7 +10,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ['3.12']
+        python-version: ['3.13']
     steps:
     - uses: actions/checkout@v2
     - name: Set up Python ${{matrix.python-version}}

pykoop/_sklearn_metaestimators/metaestimators.py

@@ -15,6 +15,7 @@ class to do so. My compromise was to copy and adjust that code for my own uses,
 """
 
 import abc
+from contextlib import suppress
 from typing import Any, Dict, List
 
 import numpy as np
@@ -28,10 +29,18 @@ def _get_params(self, attr: str, deep: bool = True) -> Dict[str, Any]:
         out = super().get_params(deep=deep)
         if not deep:
             return out
+
         estimators = getattr(self, attr)
-        if not hasattr(estimators, '__iter__'):
+        try:
+            out.update(estimators)
+        except (TypeError, ValueError):
+            # Ignore TypeError for cases where estimators is not a list of
+            # (name, estimator) and ignore ValueError when the list is not
+            # formatted correctly. This is to prevent errors when calling
+            # `set_params`. `BaseEstimator.set_params` calls `get_params` which
+            # can error for invalid values for `estimators`.
             return out
-        out.update(estimators)
+
         for name, estimator in estimators:
             if hasattr(estimator, "get_params"):
                 for key, value in estimator.get_params(deep=True).items():
@@ -43,14 +52,18 @@ def _set_params(self, attr: str, **params) -> '_BaseComposition':
         # 1. All steps
         if attr in params:
             setattr(self, attr, params.pop(attr))
-        # 2. Step replacement
+        # 2. Replace items with estimators in params
         items = getattr(self, attr)
-        names = []
-        if hasattr(items, '__iter__'):
-            names, _ = zip(*items)
-        for name in list(params.keys()):
-            if "__" not in name and name in names:
-                self._replace_estimator(attr, name, params.pop(name))
+        if isinstance(items, list) and items:
+            # Get item names used to identify valid names in params
+            # `zip` raises a TypeError when `items` does not contains
+            # elements of length 2
+            with suppress(TypeError):
+                item_names, _ = zip(*items)
+                for name in list(params.keys()):
+                    if "__" not in name and name in item_names:
+                        self._replace_estimator(attr, name, params.pop(name))
+
         # 3. Step parameters and other initialisation arguments
         super().set_params(**params)
         return self
@@ -68,11 +81,11 @@ def _validate_names(self, names: List[str]) -> None:
         if len(set(names)) != len(names):
             raise ValueError("Names provided are not unique: {0!r}".format(
                 list(names)))
-        conflict_names = set(names).intersection(self.get_params(deep=False))
-        if conflict_names:
+        invalid_names = set(names).intersection(self.get_params(deep=False))
+        if invalid_names:
             raise ValueError(
                 "Estimator names conflict with constructor arguments: {0!r}".
-                format(sorted(conflict_names)))
+                format(sorted(invalid_names)))
         invalid_names = [name for name in names if "__" in name]
         if invalid_names:
             raise ValueError(

pykoop/centers.py

@@ -78,7 +78,7 @@ class GridCenters(Centers):
     >>> grid.fit(X_msd[:, 1:])  # Remove episode feature
     GridCenters(n_points_per_feature=4)
     >>> grid.centers_
-    array([...])
+    array(...)
     """
 
     def __init__(
@@ -155,7 +155,7 @@ class UniformRandomCenters(Centers):
     >>> rand.fit(X_msd[:, 1:])  # Remove episode feature
     UniformRandomCenters(n_centers=10)
     >>> rand.centers_
-    array([...])
+    array(...)
     """
 
     def __init__(
@@ -235,7 +235,7 @@ class GaussianRandomCenters(Centers):
     >>> rand.fit(X_msd[:, 1:])  # Remove episode feature
     GaussianRandomCenters(n_centers=10)
     >>> rand.centers_
-    array([...])
+    array(...)
     """
 
     def __init__(
@@ -306,15 +306,15 @@ class QmcCenters(Centers):
     >>> qmc.fit(X_msd[:, 1:])  # Remove episode feature
     QmcCenters(n_centers=10)
     >>> qmc.centers_
-    array([...])
+    array(...)
 
     Generate centers using a Sobol sequence
 
     >>> qmc = pykoop.QmcCenters(n_centers=8, qmc=scipy.stats.qmc.Sobol)
     >>> qmc.fit(X_msd[:, 1:])  # Remove episode feature
     QmcCenters(n_centers=8, qmc=<class 'scipy.stats._qmc.Sobol'>)
     >>> qmc.centers_
-    array([...])
+    array(...)
     """
 
     def __init__(
@@ -430,7 +430,7 @@ class ClusterCenters(Centers):
     >>> kmeans.fit(X_msd[:, 1:])  # Remove episode feature
     ClusterCenters(estimator=KMeans(n_clusters=3))
     >>> kmeans.centers_
-    array([...])
+    array(...)
     """
 
     def __init__(
@@ -507,7 +507,7 @@ class GaussianMixtureRandomCenters(Centers):
     >>> gmm.fit(X_msd[:, 1:])  # Remove episode feature
     GaussianMixtureRandomCenters(estimator=GaussianMixture(n_components=3))
     >>> gmm.centers_
-    array([...])
+    array(...)
     """
 
     def __init__(

pykoop/kernel_approximation.py

@@ -113,7 +113,7 @@ class RandomFourierKernelApprox(KernelApproximation):
     >>> ka.fit(X_msd[:, 1:])  # Remove episode feature
     RandomFourierKernelApprox(n_components=10, random_state=1234)
     >>> ka.transform(X_msd[:, 1:])
-    array([...])
+    array(...)
     """
 
     _ft_lookup = {
@@ -267,6 +267,10 @@ def transform(self, X: np.ndarray) -> np.ndarray:
         """
         sklearn.utils.validation.check_is_fitted(self)
         X = sklearn.utils.validation.check_array(X)
+        if X.shape[1] != self.n_features_in_:
+            raise ValueError(
+                f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                f"is expecting {self.n_features_in_} features as input.")
         X_scaled = np.sqrt(2 * self.shape) * X
         products = X_scaled @ self.random_weights_  # (n_samples, n_components)
         if self.method == 'weight_only':
@@ -314,7 +318,7 @@ class RandomBinningKernelApprox(KernelApproximation):
     >>> ka.fit(X_msd[:, 1:])  # Remove episode feature
     RandomBinningKernelApprox(n_components=10, random_state=1234)
     >>> ka.transform(X_msd[:, 1:])
-    array([...])
+    array(...)
     """
 
     _ddot_lookup = {
@@ -451,6 +455,10 @@ def transform(self, X: np.ndarray) -> np.ndarray:
         """
         sklearn.utils.validation.check_is_fitted(self)
         X = sklearn.utils.validation.check_array(X)
+        if X.shape[1] != self.n_features_in_:
+            raise ValueError(
+                f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                f"is expecting {self.n_features_in_} features as input.")
         X_scaled = np.sqrt(2 * self.shape) * X
         X_hashed = self._hash_samples(X_scaled)
         Xt = self.encoder_.transform(X_hashed) / np.sqrt(self.n_components)
@@ -479,6 +487,6 @@ def _hash_samples(self, X: np.ndarray) -> np.ndarray:
                 hash(tuple(coords[sample, :, component].astype(int)))
                 for component in range(coords.shape[2])
             ] for sample in range(coords.shape[0])],
-            dtype=int,
+            dtype=np.int64,
         )
         return coords_hashed

pykoop/koopman_pipeline.py

@@ -721,7 +721,7 @@ def fit(
         }
         # Validate data
         X = sklearn.utils.validation.check_array(X, **self._check_array_params)
-        # Set numbre of input features (including episode feature)
+        # Set number of input features (including episode feature)
         self.n_features_in_ = X.shape[1]
         # Extract episode feature
         if self.episode_feature_:
@@ -751,12 +751,11 @@ def transform(self, X: np.ndarray) -> np.ndarray:
                 X,
                 **self._check_array_params,
             )
-            # Check input shape
+            # Check number of features
             if X.shape[1] != self.n_features_in_:
-                raise ValueError(f'{self.__class__.__name__} `fit()` called '
-                                 f'with {self.n_features_in_} features, but '
-                                 f'`transform()` called with {X.shape[1]} '
-                                 'features.')
+                raise ValueError(
+                    f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_in_} features as input.")
         return self._apply_transform_or_inverse(X, 'transform')
 
     def inverse_transform(self, X: np.ndarray) -> np.ndarray:
@@ -996,12 +995,11 @@ def transform(self, X: np.ndarray) -> np.ndarray:
                 X,
                 **self._check_array_params,
             )
-            # Check input shape
+            # Check number of features
             if X.shape[1] != self.n_features_in_:
-                raise ValueError(f'{self.__class__.__name__} `fit()` called '
-                                 f'with {self.n_features_in_} features, but '
-                                 f'`transform()` called with {X.shape[1]} '
-                                 'features.')
+                raise ValueError(
+                    f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_in_} features as input.")
         return self._apply_transform_or_inverse(X, 'transform')
 
     def inverse_transform(self, X: np.ndarray) -> np.ndarray:
@@ -1014,12 +1012,11 @@ def inverse_transform(self, X: np.ndarray) -> np.ndarray:
                 X,
                 **self._check_array_params,
             )
-            # Check input shape
+            # Check number of features
             if X.shape[1] != self.n_features_out_:
-                raise ValueError(f'{self.__class__.__name__} `fit()` output '
-                                 f'{self.n_features_out_} features, but '
-                                 '`inverse_transform()` called with '
-                                 f'{X.shape[1]} features.')
+                raise ValueError(
+                    f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_out_} features as input.")
         return self._apply_transform_or_inverse(X, 'inverse_transform')
 
     def _apply_transform_or_inverse(self, X: np.ndarray,
@@ -1270,6 +1267,11 @@ def predict(self, X: np.ndarray) -> np.ndarray:
         self._validate_feature_names(X)
         # Validate array
         X = sklearn.utils.validation.check_array(X, **self._check_array_params)
+        # Check number of features
+        if X.shape[1] != self.n_features_in_:
+            raise ValueError(
+                f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                f"is expecting {self.n_features_in_} features as input.")
         # Split episodes
         episodes = split_episodes(X, episode_feature=self.episode_feature_)
         # Predict for each episode
@@ -1623,8 +1625,13 @@ def _validate_feature_names(self, X: np.ndarray) -> None:
         if not np.all(_extract_feature_names(X) == self.feature_names_in_):
             raise ValueError('Input features do not match fit features.')
 
-    # Extra estimator tags
-    # https://scikit-learn.org/stable/developers/develop.html#estimator-tags
+    def __sklearn_tags__(self):
+        tags = super().__sklearn_tags__()
+        tags.target_tags.required = False
+        tags.target_tags.single_output = False
+        tags.target_tags.multi_output = True
+        return tags
+
     def _more_tags(self):
         return {
             'multioutput': True,
@@ -1812,12 +1819,11 @@ def transform(self, X: np.ndarray) -> np.ndarray:
                 X,
                 **self._check_array_params,
             )
-            # Check input shape
+            # Check number of features
             if X.shape[1] != self.n_features_in_:
-                raise ValueError(f'{self.__class__.__name__} `fit()` called '
-                                 f'with {self.n_features_in_} features, but '
-                                 f'`transform()` called with {X.shape[1]} '
-                                 'features.')
+                raise ValueError(
+                    f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_in_} features as input.")
         # Split episodes
         episodes = split_episodes(X, episode_feature=self.episode_feature_)
         episodes_state = []
@@ -1873,12 +1879,11 @@ def inverse_transform(self, X: np.ndarray) -> np.ndarray:
             sklearn.utils.validation.check_is_fitted(self)
             X = sklearn.utils.validation.check_array(
                 X, **self._check_array_params)
-            # Check input shape
+            # Check number of features
             if X.shape[1] != self.n_features_out_:
-                raise ValueError(f'{self.__class__.__name__} `fit()` output '
-                                 f'{self.n_features_out_} features, but '
-                                 '`inverse_transform()` called with '
-                                 f'{X.shape[1]} features.')
+                raise ValueError(
+                    f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_out_} features as input.")
         # Split episodes
         episodes = split_episodes(X, episode_feature=self.episode_feature_)
         episodes_state = []
@@ -2278,12 +2283,11 @@ def transform(self, X: np.ndarray) -> np.ndarray:
                 X,
                 **self._check_array_params,
             )
-            # Check input shape
+            # Check number of features
             if X.shape[1] != self.n_features_in_:
-                raise ValueError(f'{self.__class__.__name__} `fit()` called '
-                                 f'with {self.n_features_in_} features, but '
-                                 f'`transform()` called with {X.shape[1]} '
-                                 'features.')
+                raise ValueError(
+                    f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_in_} features as input.")
         # Apply lifting functions
         X_out = X
         for _, lf in self.lifting_functions_:
@@ -2309,12 +2313,11 @@ def inverse_transform(self, X: np.ndarray) -> np.ndarray:
                 X,
                 **self._check_array_params,
             )
-            # Check input shape
+            # Check number of features
             if X.shape[1] != self.n_features_out_:
-                raise ValueError(f'{self.__class__.__name__} `fit()` output '
-                                 f'{self.n_features_out_} features, but '
-                                 '`inverse_transform()` called with '
-                                 f'{X.shape[1]} features.')
+                raise ValueError(
+                    f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_out_} features as input.")
         # Apply inverse lifting functions in reverse order
         X_out = X
         for _, lf in self.lifting_functions_[::-1]:
@@ -2360,6 +2363,11 @@ def predict(self, X: np.ndarray) -> np.ndarray:
                 X,
                 **self._check_array_params,
             )
+            # Check number of features
+            if X.shape[1] != self.n_features_in_:
+                raise ValueError(
+                    f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                    f"is expecting {self.n_features_in_} features as input.")
         # Lift data matrix
         X_trans = self.transform(X)
         # Predict in lifted space
@@ -2405,6 +2413,11 @@ def score(self, X: np.ndarray, y: Optional[np.ndarray] = None) -> float:
         self._validate_feature_names(X)
         # Validate input array
         X = sklearn.utils.validation.check_array(X, **self._check_array_params)
+        # Check number of features
+        if X.shape[1] != self.n_features_in_:
+            raise ValueError(
+                f"X has {X.shape[1]} features, but {self.__class__.__name__} "
+                f"is expecting {self.n_features_in_} features as input.")
         scorer = KoopmanPipeline.make_scorer()
         score = scorer(self, X, None)
         return score

pykoop/lifting_functions.py

@@ -72,9 +72,9 @@ class SkLearnLiftingFn(koopman_pipeline.EpisodeIndependentLiftingFn):
     ['ep', 'StandardScaler(x0)', 'StandardScaler(x1)', 'StandardScaler(u0)']
     >>> X_msd_pp = std_scaler.transform(X_msd)
     >>> np.mean(X_msd_pp[:, 1:], axis=0)
-    array([...])
+    array(...)
     >>> np.std(X_msd_pp[:, 1:], axis=0)
-    array([...])
+    array(...)
     """
 
     def __init__(