add softmax activation for entity embedding

AxeldeRomblay · Jan 28, 2019 · 8e576df · 8e576df
1 parent e18dafe
commit 8e576df
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/python-package/mlbox/encoding/categorical_encoder.py b/python-package/mlbox/encoding/categorical_encoder.py
@@ -235,7 +235,7 @@ def fit(self, df_train, y_train):
 
                         outputs = Dense(y_train.nunique(),
                                         kernel_initializer='normal',
-                                        activation='sigmoid')(lay2)
+                                        activation='softmax')(lay2)
 
                         model = Model(inputs=inputs, outputs=outputs)
                         model.compile(loss='binary_crossentropy',