scikit-learn · jeremiedbb · Jun 28, 2022 · May 31, 2022 · Jun 2, 2022 · Jun 3, 2022
diff --git a/sklearn/dummy.py b/sklearn/dummy.py
@@ -4,13 +4,16 @@
 # License: BSD 3 clause
 
 import warnings
+from numbers import Integral, Real
+
 import numpy as np
 import scipy.sparse as sp
 
 from .base import BaseEstimator, ClassifierMixin, RegressorMixin
 from .base import MultiOutputMixin
 from .utils import check_random_state
 from .utils import deprecated
+from .utils._param_validation import StrOptions, Interval
 from .utils.validation import _num_samples
 from .utils.validation import check_array
 from .utils.validation import check_consistent_length
@@ -134,6 +137,14 @@ class prior probabilities.
     0.75
     """
 
+    _parameter_constraints = {
+        "strategy": [
+            StrOptions({"most_frequent", "prior", "stratified", "uniform", "constant"})
+        ],
+        "random_state": ["random_state"],
+        "constant": [Integral, str, "array-like", None],
+    }
+
     def __init__(self, *, strategy="prior", random_state=None, constant=None):
         self.strategy = strategy
         self.random_state = random_state
@@ -158,19 +169,7 @@ def fit(self, X, y, sample_weight=None):
         self : object
             Returns the instance itself.
         """
-        allowed_strategies = (
-            "most_frequent",
-            "stratified",
-            "uniform",
-            "constant",
-            "prior",
-        )
-
-        if self.strategy not in allowed_strategies:
-            raise ValueError(
-                "Unknown strategy type: %s, expected one of %s."
-                % (self.strategy, allowed_strategies)
-            )
+        self._validate_params()
 
         self._strategy = self.strategy
 
@@ -527,6 +526,16 @@ class DummyRegressor(MultiOutputMixin, RegressorMixin, BaseEstimator):
     0.0
     """
 
+    _parameter_constraints = {
+        "strategy": [StrOptions({"mean", "median", "quantile", "constant"})],
+        "quantile": [Interval(Real, 0.0, 1.0, closed="both"), None],
+        "constant": [
+            Interval(Real, None, None, closed="neither"),
+            "array-like",
+            None,
+        ],
+    }
+
     def __init__(self, *, strategy="mean", constant=None, quantile=None):
         self.strategy = strategy
         self.constant = constant
@@ -551,12 +560,7 @@ def fit(self, X, y, sample_weight=None):
         self : object
             Fitted estimator.
         """
-        allowed_strategies = ("mean", "median", "quantile", "constant")
-        if self.strategy not in allowed_strategies:
-            raise ValueError(
-                "Unknown strategy type: %s, expected one of %s."
-                % (self.strategy, allowed_strategies)
-            )
+        self._validate_params()
 
         y = check_array(y, ensure_2d=False, input_name="y")
         if len(y) == 0:
@@ -584,12 +588,11 @@ def fit(self, X, y, sample_weight=None):
                 ]
 
         elif self.strategy == "quantile":
-            if self.quantile is None or not np.isscalar(self.quantile):
+            if self.quantile is None:
                 raise ValueError(
-                    "Quantile must be a scalar in the range [0.0, 1.0], but got %s."
-                    % self.quantile
+                    "When using `strategy='quantile', you have to specify the desired "
+                    "quantile in the range [0, 1]."
                 )
-
             percentile = self.quantile * 100.0
             if sample_weight is None:
                 self.constant_ = np.percentile(y, axis=0, q=percentile)

diff --git a/sklearn/model_selection/tests/test_successive_halving.py b/sklearn/model_selection/tests/test_successive_halving.py
@@ -29,6 +29,15 @@ class FastClassifier(DummyClassifier):
     These parameter don't affect the predictions and are useful for fast
     grid searching."""
 
+    # update the constraints such that we accept all parameters from a to z
+    _parameter_constraints = {
+        **DummyClassifier._parameter_constraints,
+        **{
+            chr(key): "no_validation"  # type: ignore
+            for key in range(ord("a"), ord("z") + 1)
+        },
+    }
+
     def __init__(
         self, strategy="stratified", random_state=None, constant=None, **kwargs
     ):

diff --git a/sklearn/tests/test_common.py b/sklearn/tests/test_common.py
@@ -468,8 +468,6 @@ def test_estimators_do_not_raise_errors_in_init_or_set_params(Estimator):
     "DBSCAN",
     "DictVectorizer",
     "DictionaryLearning",
-    "DummyClassifier",
-    "DummyRegressor",
     "ElasticNet",
     "ElasticNetCV",
     "EllipticEnvelope",

diff --git a/sklearn/tests/test_dummy.py b/sklearn/tests/test_dummy.py
@@ -231,17 +231,6 @@ def test_classifier_prediction_independent_of_X(strategy):
     assert_array_equal(predictions1, predictions2)
 
 
-def test_classifier_exceptions():
-    clf = DummyClassifier(strategy="unknown")
-    with pytest.raises(ValueError):
-        clf.fit([], [])
-
-    with pytest.raises(NotFittedError):
-        clf.predict([])
-    with pytest.raises(NotFittedError):
-        clf.predict_proba([])
-
-
 def test_mean_strategy_regressor():
 
     random_state = np.random.RandomState(seed=1)
@@ -379,28 +368,11 @@ def test_quantile_invalid():
     X = [[0]] * 5  # ignored
     y = [0] * 5  # ignored
 
-    est = DummyRegressor(strategy="quantile")
-    with pytest.raises(ValueError):
-        est.fit(X, y)
-
     est = DummyRegressor(strategy="quantile", quantile=None)
-    with pytest.raises(ValueError):
-        est.fit(X, y)
-
-    est = DummyRegressor(strategy="quantile", quantile=[0])
-    with pytest.raises(ValueError):
-        est.fit(X, y)
-
-    est = DummyRegressor(strategy="quantile", quantile=-0.1)
-    with pytest.raises(ValueError):
-        est.fit(X, y)
-
-    est = DummyRegressor(strategy="quantile", quantile=1.1)
-    with pytest.raises(ValueError):
-        est.fit(X, y)
-
-    est = DummyRegressor(strategy="quantile", quantile="abc")
-    with pytest.raises(TypeError):
+    err_msg = (
+        "When using `strategy='quantile', you have to specify the desired quantile"
+    )
+    with pytest.raises(ValueError, match=err_msg):
         est.fit(X, y)
 
 
@@ -462,21 +434,13 @@ def test_y_mean_attribute_regressor():
     assert est.constant_ == np.mean(y)
 
 
-def test_unknown_strategey_regressor():
-    X = [[0]] * 5
-    y = [1, 2, 4, 6, 8]
-
-    est = DummyRegressor(strategy="gona")
-    with pytest.raises(ValueError):
-        est.fit(X, y)
-
-
 def test_constants_not_specified_regressor():
     X = [[0]] * 5
     y = [1, 2, 4, 6, 8]
 
     est = DummyRegressor(strategy="constant")
-    with pytest.raises(TypeError):
+    err_msg = "Constant target value has to be specified"
+    with pytest.raises(TypeError, match=err_msg):
         est.fit(X, y)
 
 
@@ -486,7 +450,8 @@ def test_constant_size_multioutput_regressor():
     y = random_state.randn(10, 5)
 
     est = DummyRegressor(strategy="constant", constant=[1, 2, 3, 4])
-    with pytest.raises(ValueError):
+    err_msg = r"Constant target value should have shape \(5, 1\)."
+    with pytest.raises(ValueError, match=err_msg):
         est.fit(X, y)
 
 
@@ -554,7 +519,6 @@ def test_constant_strategy_exceptions(y, params, err_msg):
     X = [[0], [0], [0], [0]]
 
     clf = DummyClassifier(strategy="constant", **params)
-
     with pytest.raises(ValueError, match=err_msg):
         clf.fit(X, y)