Generative Deep Learning

Chapter 2. Deep Learning

Let’s start with a basic definition of deep learning:

Deep learning is a class of machine learning algorithm that uses multiple stacked layers of processing units to learn high-level representations from unstructured data.

To understand deep learning fully, and particularly why it is so useful within generative modeling, we need to delve into this definition a bit further. First, what do we mean by “unstructured data” and its counterpart, “structured data”?

Structured and Unstructured Data

Many types of machine learning algorithm require structured, tabular data as input, arranged into columns of features that describe each observation. For example, a person’s age, income, and number of website visits in the last month are all features that could help to predict if the person will subscribe to a particular online service in the coming month. We could use a structured table of these features to train a logistic regression, random forest, or XGBoost model to predict the binary response variable—did the person subscribe (1) or not (0)? Here, each individual feature contains a nugget of information about the observation, and the model would learn how these features interact to influence the response.

Unstructured data refers to any data that is not naturally arranged into columns of features, such as images, audio, and text. There is of course spatial structure to an image, temporal structure to a recording, and both spatial and temporal structure to video ...