GRU with Keras

An advantage of using TensorFlow and Keras is that they make it easy to create models. Just like LSTM, creating a GRU model is only a matter of adding the GRU layer instead of LSTM or SimpleRNN layer, as follows:

model.add(GRU(units=4, input_shape=(X_train.shape[1], X_train.shape[2])))

The model structure is as follows:

Layer (type)                 Output Shape              Param #   
=================================================================
gru_1 (GRU)                  (None, 4)                 72        
_________________________________________________________________
dense_1 (Dense)              (None, 1)                 5         
=================================================================
Total params: 77
Trainable params: 77
Non-trainable params: 0

The complete code for the GRU model is provided in notebook ch-07b_RNN_TimeSeries_Keras.

As expected...