Chapter 11. RNN with TensorFlow and Keras

In problems involving ordered sequences of data, such as time series Forecasting and natural language processing, the context is very valuable to predict the output. The context for such problems can be determined by ingesting the whole sequence, not just one last data point. Thus, the previous output becomes part of the current input, and when repeated, the last output turns out to be the results of all the previous inputs along with the last input. Recurrent Neural Network (RNN) architecture is a solution for handling  machine learning problems that involve sequences.

Recurrent Neural Network (RNN) is a specialized neural network architecture for handling sequential data. The sequential data could be the sequence of observations over a period of time, as in time series data, or sequence of characters, words, and sentences, as in textual data.

One of the assumptions for the standard neural network is that the input data is arranged in a way that one...