Update _Dist/NeuralNetworks (beta)

Author: carefree0910

_Dist/NeuralNetworks/e_AdvancedNN/NN.py

@@ -53,16 +53,16 @@ def init_data_info(self):
             raise ValueError("numerical_idx should be provided")
         if self.categorical_columns is None:
             raise ValueError("categorical_columns should be provided")
+
+    def init_from_data(self, x, y, x_test, y_test, sample_weights, names):
+        self.init_data_info()
+        super(Advanced, self).init_from_data(x, y, x_test, y_test, sample_weights, names)
         if len(self.numerical_idx) != self.n_dim + 1:
             raise ValueError("Length of numerical_idx should be {}, {} found".format(
                 self.n_dim + 1, len(self.numerical_idx)
             ))
         self.n_dim -= len(self.categorical_columns)
 
-    def init_from_data(self, x, y, x_test, y_test, sample_weights, names):
-        self.init_data_info()
-        super(Advanced, self).init_from_data(x, y, x_test, y_test, sample_weights, names)
-
     def init_model_param_settings(self):
         super(Advanced, self).init_model_param_settings()
         self.dropout_keep_prob = self.model_param_settings.get("p_keep", 0.5)

_Dist/NeuralNetworks/f_AutoNN/NN.py

@@ -125,7 +125,6 @@ def init_from_data(self, x, y, x_test, y_test, sample_weights, names):
         self._handle_unbalance(y)
         self._handle_sparsity()
         super(Auto, self).init_from_data(x, y, x_test, y_test, sample_weights, names)
-        self.n_dim -= len(self.categorical_columns)
 
     def _handle_unbalance(self, y):
         class_ratio = self.class_prior.min() / self.class_prior.max()