Understanding Sequential Models

A sequence works in a way a collection never can.

—George Murray

This chapter covers an important class of machine learning models, the sequential models. A defining characteristic of such models is that the processing layers are arranged in such a way that the output of one layer is the input to the other. This architecture makes them perfect to process sequential data. Sequential data is the type of data that consists of ordered series of elements such as a sentence in a document or a time series of stock market prices.

In this chapter, we will start with understanding the characteristics of sequential data. Then, we will present the working of RNNs and how they can be used to process sequential data. Next, we will learn how we can address the limitations of RNN through GRU without scarifying accuracy. Then, we will discuss the architecture of LSTM. Finally, we will compare different sequential modeling architectures...