Machine Learning for Streaming Data with Python

Chapter 8: Reinforcement Learning

The reinforcement learning paradigm is very different than standard machine learning and even the online machine learning methods that we have covered in earlier chapters. Although reinforcement learning will not always be a better choice than "regular" learning for many use cases, it is a powerful tool for tackling re-learning and the adaptation of models.

In reinforcement learning, we give the model a lot of decisive power to do its re-learning and to update the rules of its decision-making process. Rather than letting the model make a prediction and hardcode the action to take for this prediction, the model will directly decide on the action to take.

For automated machine learning pipelines in which actions ...