The code for compiling the model is as follows: 

# Compile model
model %>% compile(optimizer = "rmsprop",
         loss = "binary_crossentropy",
         metrics = c("acc"))

We compile the model with the rmsprop optimizer, which is recommended for recurrent neural networks. We make use of binary_crossentropy as the loss function due to a binary type of response since movie reviews are either positive or negative. Finally, for metrics, we have specified accuracy.

In the next section, we will use this architecture to develop a movie review sentiment classification model that uses recurrent neural networks.