make inverse_link the identity by default for API consistency

Author: floriankozikowski

skglm/datafits/base.py

@@ -2,6 +2,22 @@
 class BaseDatafit:
     """Base class for datafits."""
 
+    @staticmethod
+    def inverse_link(x):
+        """Inverse link function (identity by default).
+
+        Parameters
+        ----------
+        x : array-like
+            Linear predictor values.
+
+        Returns
+        -------
+        array-like
+            Transformed values in response scale.
+        """
+        return x
+
     def get_spec(self):
         """Specify the numba types of the class attributes.
 

skglm/estimators.py

@@ -266,10 +266,8 @@ def predict(self, X):
             else:
                 indices = scores.argmax(axis=1)
             return self.classes_[indices]
-        elif hasattr(self.datafit, "inverse_link"):
-            return self.datafit.inverse_link(self._decision_function(X))
         else:
-            return self._decision_function(X)
+            return self.datafit.inverse_link(self._decision_function(X))
 
     def get_params(self, deep=False):
         """Get parameters of the estimators including the datafit's and penalty's.