Don't automatically compute feature correlations; this is a performance enhancement

Author: jonathanronen

maui/model.py

@@ -186,9 +186,8 @@ def transform(self, X, encoder="mean"):
             index=self.x_.index,
             columns=[f"LF{i}" for i in range(1, self.n_latent + 1)],
         )
-        self.feature_correlations = maui.utils.correlate_factors_and_features(
-            self.z_, self.x_
-        )
+
+        self.feature_correlations_ = None
         self.w_ = None
         return self.z_
 
@@ -447,6 +446,24 @@ def get_linear_weights(self):
             )
         return self.w_
 
+    def get_feature_correlations(self):
+        """Get correlation coefficients between input features and latent factors.
+
+        Returns
+        -------
+        r:  (n_features, n_latent_factors) DataFrame
+            r_{ij} is the correlation coefficient between feature `i`
+            and latent factor `j`.
+        """
+        if (
+            not hasattr(self, "feature_correlations_")
+            or self.feature_correlations_ is None
+        ):
+            self.feature_correlations_ = maui.utils.correlate_factors_and_features(
+                self.z_, self.x_
+            )
+        return self.feature_correlations_
+
     def drop_unexplanatory_factors(self, threshold=0.02):
         """Drops factors which have a low R^2 score in a univariate linear model
         predicting the features `x` from a column of the latent factors `z`.

test/test_maui.py

@@ -67,7 +67,9 @@ def test_dict2array():
 def test_maui_saves_feature_correlations():
     maui_model = Maui(n_hidden=[10], n_latent=2, epochs=1)
     z = maui_model.fit_transform({"d1": df1, "d2": df2})
-    assert hasattr(maui_model, "feature_correlations")
+    r = maui_model.get_feature_correlations()
+    assert r is not None
+    assert hasattr(maui_model, "feature_correlations_")
 
 
 def test_maui_saves_w():