The Bandit and Reinforcement Learning

In the previous chapter, we introduced Machine Learning and the types of learning or training used in ML (Unsupervised Training, Supervised Training, Reinforcement Learning, Imitation Learning, and Curriculum Learning). As we discussed, the various forms of learning each have their own advantages and disadvantages. While ML using supervised training has been used successfully in games as far back as 20 years ago, it never really found any traction. It wasn't until the successful use of Reinforcement Learning was shown to be capable of playing classic Atari games and GO better than humans, that the interest for ML in games and simulations was rekindled. Now, RL is one of the hottest topics in ML research and is showing the potential for building some real continually learning AI. We will spend the bulk of this chapter understanding RL...