In this chapter, we introduced some of the fundamental themes of DL. DL consists of a set of methods that allow an ML system to obtain a hierarchical representation of data on multiple levels. This is achieved by combining simple units, each of which transforms the representation at its own level, starting from the input level, in a representation at a higher and abstraction level.

Recently, these techniques have provided results that have never been seen before in many applications, such as image recognition and speech recognition. One of the main reasons for the spread of these techniques has been the development of GPU architectures that considerably reduce the training time of DNNs.

There are different DNN architectures, each of which has been developed for a specific problem. We will talk more about these architectures in later chapters and show examples of applications created with the TensorFlow framework. This chapter ended with a brief overview of the most important DL frameworks.

In the next chapter, we begin our journey into DL, introducing the TensorFlow software library. We will describe the main features of TensorFlow and see how to install it and set up our first working remarketing dataset.