5. Deep Learning for Sequences

Overview

In this chapter, we will implement deep learning-based approaches to sequence modeling, after understanding the considerations of dealing with sequences. We will begin with Recurrent Neural Networks (RNNs), an intuitive approach to sequence processing that has provided state-of-the-art results. We will then discuss and implement 1D convolutions as another approach and see how it compares with RNNs. We will also combine RNNs with 1D convolutions in a hybrid model. We will employ all of these models on a classic sequence processing task – stock price prediction. By the end of this chapter, you will become adept at implementing deep learning approaches for sequences, particularly plain RNNs and 1D convolutions, and you will have laid the foundations for more advanced RNN-based models.