Merge pull request #6534 from noahnovsak/dask-pca-fix

PCA: distinguish between PCA and IncrementalPCA

Orange/projection/pca.py

@@ -11,6 +11,11 @@
 from sklearn.utils.extmath import svd_flip, safe_sparse_dot
 from sklearn.utils.validation import check_is_fitted
 
+try:
+    import dask_ml.decomposition as dask_decomposition
+except ImportError:
+    dask_decomposition = skl_decomposition
+
 import Orange.data
 from Orange.statistics import util as ut
 from Orange.data import Variable
@@ -275,18 +280,13 @@ def _initialize_wrapped(self, X=None, Y=None):
             params["n_components"] = min(*X.shape, params["n_components"])
 
         if isinstance(X, da.Array) or isinstance(Y, da.Array):
-            try:
-                import dask_ml.decomposition as dask_decomposition
-
+            if dask_decomposition is skl_decomposition:
+                warnings.warn("dask_ml is not installed. Using sklearn instead.")
+            else:
                 if params["iterated_power"] == "auto":
                     params["iterated_power"] = 0
-                del params["tol"]
 
-                # use IPCA instead of PCA due to memory issues
-                return dask_decomposition.IncrementalPCA(**params)
-
-            except ImportError:
-                warnings.warn("dask_ml is not installed. Using sklearn instead.")
+                return dask_decomposition.PCA(**params)
 
         return self.__wraps__(**params)
 
@@ -340,8 +340,16 @@ def __init__(self, n_components=None, whiten=False, copy=True,
         super().__init__(preprocessors=preprocessors)
         self.params = vars()
 
+    def _initialize_wrapped(self, X=None, Y=None):
+        if isinstance(X, da.Array) or isinstance(Y, da.Array):
+            if dask_decomposition is skl_decomposition:
+                warnings.warn("dask_ml is not installed. Using sklearn instead.")
+            else:
+                return dask_decomposition.IncrementalPCA(**self.params)
+        return self.__wraps__(**self.params)
+
     def fit(self, X, Y=None):
-        proj = self.__wraps__(**self.params)
+        proj = self._initialize_wrapped(X, Y)
         proj = proj.fit(X, Y)
         return IncrementalPCAModel(proj, self.domain, len(proj.components_))
 

Orange/tests/test_pca.py

@@ -159,8 +159,9 @@ def test_improved_randomized_pca_sparse_data(self):
 
     @unittest.skipIf(sklearn_version.startswith('0.20'),
                      "https://github.com/scikit-learn/scikit-learn/issues/12234")
-    def test_incremental_pca(self):
-        data = self.ionosphere
+    @with_dasktable
+    def test_incremental_pca(self, prepare_table):
+        data = prepare_table(self.ionosphere)
         self.__ipca_test_helper(data, n_com=3, min_xpl_var=0.49)
         self.__ipca_test_helper(data, n_com=32, min_xpl_var=1)
 
@@ -172,7 +173,7 @@ def __ipca_test_helper(self, data, n_com, min_xpl_var):
         self.assertEqual(n_com, pca_model.n_components)
         self.assertEqual((n_com, data.X.shape[1]), pca_model.components_.shape)
         proj = np.dot(data.X - pca_model.mean_, pca_model.components_.T)
-        np.testing.assert_almost_equal(pca_model(data).X, proj)
+        np.testing.assert_almost_equal(pca_model(data).X, np.asarray(proj))
         pc1_ipca = pca_model.components_[0]
         self.assertAlmostEqual(np.linalg.norm(pc1_ipca), 1)
         pc1_pca = PCA(n_components=n_com)(data).components_[0]