Open JJ opened 8 years ago
Not for using on time-series... :/
In any case, I will test some prediction methods that a student used in his TFG.
What about using the "LSTM Recurrent Neural Network" included in Keras? That model seem to provide much better results than a simple MLP.
Do we have a viable version of G-Prop?