In the previous chapters, we covered several topics related to machine learning and focused on supervised learning tasks. While supervised learning is immensely important and has a lot of real-life applications, reinforcement learning currently seems to be the most exciting and promising branch of machine learning. The reasons for this excitement include the complex, everyday-life-like tasks that reinforcement learning has the potential to handle. In March 2016, AlphaGo, a reinforcement learning-based system specializing in the highly complex game of Go, was able to defeat the person considered to be the greatest Go player of the past decade in a competition that was widely covered by the media.
While supervised learning requires labeled data for training—in other words, pairs of inputs and matching outputs—reinforcement learning does not present...
