Defining reinforcement learning

Reinforcement learning is a subdomain of machine learning that focuses on creating machine learning models that make decisions. Sometimes, the models are not referred to as models, but rather as intelligent agents.

When looking from a distance, you could argue that reinforcement learning is very close to machine learning. We could say that both of them are methods inside artificial intelligence that try to deliver intelligent black boxes, which are able to learn specific tasks just like a human would – often better.

If we look closer, however, we start to see important differences. In previous chapters, you have seen machine learning models such as anomaly detection, classification, and regression. All of them use a number of variables and are able to make real-time predictions on a target variable based on those.

You have seen a number of metrics that allow us data scientists to decide whether a model is any good. The online models...