TST activate common tests for TSNE (scikit-learn#25374)

glemaitre · Itay · commit 467d7b537293 · 2023-04-15T23:06:51.000+03:00
diff --git a/sklearn/manifold/_t_sne.py b/sklearn/manifold/_t_sne.py
@@ -17,7 +17,7 @@
 from scipy.sparse import csr_matrix, issparse
 from numbers import Integral, Real
 from ..neighbors import NearestNeighbors
-from ..base import BaseEstimator
+from ..base import BaseEstimator, ClassNamePrefixFeaturesOutMixin, TransformerMixin
 from ..utils import check_random_state
 from ..utils._openmp_helpers import _openmp_effective_n_threads
 from ..utils.validation import check_non_negative
@@ -537,7 +537,7 @@ def trustworthiness(X, X_embedded, *, n_neighbors=5, metric="euclidean"):
     return t
 
 
-class TSNE(BaseEstimator):
+class TSNE(ClassNamePrefixFeaturesOutMixin, TransformerMixin, BaseEstimator):
     """T-distributed Stochastic Neighbor Embedding.
 
     t-SNE [1] is a tool to visualize high-dimensional data. It converts
@@ -1145,5 +1145,10 @@ def fit(self, X, y=None):
         self.fit_transform(X)
         return self
 
+    @property
+    def _n_features_out(self):
+        """Number of transformed output features."""
+        return self.embedding_.shape[1]
+
     def _more_tags(self):
         return {"pairwise": self.metric == "precomputed"}
diff --git a/sklearn/utils/estimator_checks.py b/sklearn/utils/estimator_checks.py
@@ -4234,9 +4234,14 @@ def fit_then_transform(est):
     def fit_transform(est):
         return est.fit_transform(X, y)
 
-    transform_methods = [fit_then_transform, fit_transform]
-    for transform_method in transform_methods:
+    transform_methods = {
+        "transform": fit_then_transform,
+        "fit_transform": fit_transform,
+    }
+    for name, transform_method in transform_methods.items():
         transformer = clone(transformer)
+        if not hasattr(transformer, name):
+            continue
         X_trans_no_setting = transform_method(transformer)
 
         # Auto wrapping only wraps the first array
@@ -4269,29 +4274,31 @@ def _output_from_fit_transform(transformer, name, X, df, y):
         ("fit.transform/array/df", X, df),
         ("fit.transform/array/array", X, X),
     ]
-    for (
-        case,
-        data_fit,
-        data_transform,
-    ) in cases:
-        transformer.fit(data_fit, y)
-        if name in CROSS_DECOMPOSITION:
-            X_trans, _ = transformer.transform(data_transform, y)
-        else:
-            X_trans = transformer.transform(data_transform)
-        outputs[case] = (X_trans, transformer.get_feature_names_out())
+    if all(hasattr(transformer, meth) for meth in ["fit", "transform"]):
+        for (
+            case,
+            data_fit,
+            data_transform,
+        ) in cases:
+            transformer.fit(data_fit, y)
+            if name in CROSS_DECOMPOSITION:
+                X_trans, _ = transformer.transform(data_transform, y)
+            else:
+                X_trans = transformer.transform(data_transform)
+            outputs[case] = (X_trans, transformer.get_feature_names_out())
 
     # fit_transform case:
     cases = [
         ("fit_transform/df", df),
         ("fit_transform/array", X),
     ]
-    for case, data in cases:
-        if name in CROSS_DECOMPOSITION:
-            X_trans, _ = transformer.fit_transform(data, y)
-        else:
-            X_trans = transformer.fit_transform(data, y)
-        outputs[case] = (X_trans, transformer.get_feature_names_out())
+    if hasattr(transformer, "fit_transform"):
+        for case, data in cases:
+            if name in CROSS_DECOMPOSITION:
+                X_trans, _ = transformer.fit_transform(data, y)
+            else:
+                X_trans = transformer.fit_transform(data, y)
+            outputs[case] = (X_trans, transformer.get_feature_names_out())
 
     return outputs
 
