FIX convert cv_results_ values to numpy array in SuccessiveHalving (#19211)

NicolasHug · thomasjpfan · web-flow · commit 872052b9ab47 · 2021-04-14T09:30:25.000+02:00
Co-authored-by: Thomas J. Fan &lt;thomasjpfan@gmail.com&gt;
diff --git a/doc/whats_new/v0.24.rst b/doc/whats_new/v0.24.rst
@@ -40,6 +40,11 @@ Changelog
   :class:`model_selection.GridSearchCV` now correctly shows the score for
   single metrics and verbose > 2. :pr:`19659` by `Thomas Fan`_.
 
+- |Fix| Some values in the `cv_results_` attribute of
+  :class:`model_selection.HalvingRandomSearchCV` and
+  :class:`model_selection.HalvingGridSearchCV` were not properly converted to
+  numpy arrays. :pr:`19211` by `Nicolas Hug`_.
+
 :mod:`sklearn.preprocessing`
 ............................
 
diff --git a/sklearn/model_selection/_search.py b/sklearn/model_selection/_search.py
@@ -897,6 +897,10 @@ def _format_results(self, candidate_params, n_splits, out,
         out = _aggregate_score_dicts(out)
 
         results = dict(more_results or {})
+        for key, val in results.items():
+            # each value is a list (as per evaluate_candidate's convention)
+            # we convert it to an array for consistency with the other keys
+            results[key] = np.asarray(val)
 
         def _store(key_name, array, weights=None, splits=False, rank=False):
             """A small helper to store the scores/times to the cv_results_"""
diff --git a/sklearn/model_selection/tests/test_successive_halving.py b/sklearn/model_selection/tests/test_successive_halving.py
@@ -445,6 +445,12 @@ def scorer(est, X, y):
         sh.set_params(n_candidates=2 * 30, min_resources='exhaust')
 
     sh.fit(X, y)
+
+    # non-regression check for
+    # https://github.com/scikit-learn/scikit-learn/issues/19203
+    assert isinstance(sh.cv_results_['iter'], np.ndarray)
+    assert isinstance(sh.cv_results_['n_resources'], np.ndarray)
+
     cv_results_df = pd.DataFrame(sh.cv_results_)
 
     # just make sure we don't have ties
