In this chapter, we've looked at how to use recurrent neural networks to make predictions based on time series data. Recurrent neural networks are useful in scenarios where you have to deal with financial data, IoT data, or any other information that is collected over time.
One important building block for recurrent neural networks is the Fold and the Recurrence layer types, which you can combine with any of the recurrent layer types, such as RNNStep, GRU, or LSTM, to build a recurrent layer set. Depending on whether you want to predict a sequence or single value, you can use the Recurrence or Fold layer types to wrap the recurrent layers.
When you're training a recurrent neural network, you can make use of the sequence data stored in the CTF file format to make it easier to train the model. But, you can just as easily use sequences stored as numpy arrays,...
