ENH Allow isotonic reg 2d input with 1 feature (GH15012)

doc/whats_new/v0.24.rst

@@ -61,6 +61,13 @@ Changelog
   change since `None` was defaulting to these values already.
   :pr:`16493` by :user:`Darshan N <DarshanGowda0>`.
 
+:mod:`sklearn.isotonic`
+.......................
+
+- |Enhancement| :meth:`IsotonicRegression.fit`, :meth:`IsotonicRegression.transform`
+  and :meth:`IsotonicRegression.predict` now accepts 2darray with 1 feature as
+  input array. :pr:`17379` by :user:`Jiaxiang <fujiaxiang>`.
+
 :mod:`sklearn.metrics`
 ......................
 

sklearn/isotonic.py

@@ -199,7 +199,7 @@ class IsotonicRegression(RegressorMixin, TransformerMixin, BaseEstimator):
     >>> from sklearn.datasets import make_regression
     >>> from sklearn.isotonic import IsotonicRegression
     >>> X, y = make_regression(n_samples=10, n_features=1, random_state=41)
-    >>> iso_reg = IsotonicRegression().fit(X.flatten(), y)
+    >>> iso_reg = IsotonicRegression().fit(X, y)
     >>> iso_reg.predict([.1, .2])
     array([1.8628..., 3.7256...])
     """
@@ -211,9 +211,11 @@ def __init__(self, *, y_min=None, y_max=None, increasing=True,
         self.increasing = increasing
         self.out_of_bounds = out_of_bounds
 
-    def _check_fit_data(self, X, y, sample_weight=None):
-        if len(X.shape) != 1:
-            raise ValueError("X should be a 1d array")
+    def _check_input_data_shape(self, X, y=None, sample_weight=None):
+        if not (len(X.shape) == 1 or (len(X.shape) == 2 and X.shape[1] == 1)):
+            msg = "Isotonic regression input X should be a 1d array or " \
+                  "2d array with 1 feature"
+            raise ValueError(msg)
 
     def _build_f(self, X, y):
         """Build the f_ interp1d function."""
@@ -234,7 +236,8 @@ def _build_f(self, X, y):
 
     def _build_y(self, X, y, sample_weight, trim_duplicates=True):
         """Build the y_ IsotonicRegression."""
-        self._check_fit_data(X, y, sample_weight)
+        self._check_input_data_shape(X, y, sample_weight)
+        X = X.reshape(-1)
 
         # Determine increasing if auto-determination requested
         if self.increasing == 'auto':
@@ -283,7 +286,7 @@ def fit(self, X, y, sample_weight=None):
 
         Parameters
         ----------
-        X : array-like of shape (n_samples,)
+        X : array-like of shape (n_samples,) or (n_samples, 1)
             Training data.
 
         y : array-like of shape (n_samples,)
@@ -327,7 +330,7 @@ def transform(self, T):
 
         Parameters
         ----------
-        T : array-like of shape (n_samples,)
+        T : array-like of shape (n_samples,) or (n_samples, 1)
             Data to transform.
 
         Returns
@@ -343,8 +346,8 @@ def transform(self, T):
 
         T = check_array(T, dtype=dtype, ensure_2d=False)
 
-        if len(T.shape) != 1:
-            raise ValueError("Isotonic regression input should be a 1d array")
+        self._check_input_data_shape(T)
+        T = T.reshape(-1)
 
         # Handle the out_of_bounds argument by clipping if needed
         if self.out_of_bounds not in ["raise", "nan", "clip"]:
@@ -367,7 +370,7 @@ def predict(self, T):
 
         Parameters
         ----------
-        T : array-like of shape (n_samples,)
+        T : array-like of shape (n_samples,) or (n_samples, 1)
             Data to transform.
 
         Returns

sklearn/tests/test_isotonic.py

@@ -9,9 +9,10 @@
                               IsotonicRegression, _make_unique)
 
 from sklearn.utils.validation import check_array
-from sklearn.utils._testing import (assert_raises, assert_array_equal,
-                                   assert_array_almost_equal,
-                                   assert_warns_message, assert_no_warnings)
+from sklearn.utils._testing import (assert_raises, assert_allclose,
+                                    assert_array_equal,
+                                    assert_array_almost_equal,
+                                    assert_warns_message, assert_no_warnings)
 from sklearn.utils import shuffle
 
 from scipy.special import expit
@@ -508,3 +509,36 @@ def test_make_unique_dtype():
         w = np.ones_like(x)
         x, y, w = _make_unique(x, y, w)
         assert_array_equal(x, [2, 3, 5])
+
+
+def test_input_shape_validation():
+    # Test from #15012
+    # Check that IsotonicRegression can handle 2darray with only 1 feature
+    X = np.arange(10)
+    X_2d = X.reshape(-1, 1)
+    y = np.arange(10)
+
+    iso_reg = IsotonicRegression().fit(X, y)
+    iso_reg_2d = IsotonicRegression().fit(X_2d, y)
+
+    y_pred1 = iso_reg.predict(X)
+    y_pred2 = iso_reg_2d.predict(X_2d)
+    assert_allclose(y_pred1, y_pred2)
+
+
+def test_isotonic_2darray_more_than_1_feature():
+    # Ensure IsotonicRegression raises error if input has more than 1 feature
+    X = np.arange(10)
+    X_2d = np.c_[X, X]
+    y = np.arange(10)
+
+    msg = "should be a 1d array or 2d array with 1 feature"
+    with pytest.raises(ValueError, match=msg):
+        IsotonicRegression().fit(X_2d, y)
+
+    iso_reg = IsotonicRegression().fit(X, y)
+    with pytest.raises(ValueError, match=msg):
+        iso_reg.predict(X_2d)
+
+    with pytest.raises(ValueError, match=msg):
+        iso_reg.transform(X_2d)