In this chapter, we have learned about the MAB problem and how it can be applied to different applications. We understood several methods to solve an explore-exploit dilemma. First, we looked at the epsilon-greedy policy, where we explored with the probability epsilon, and carried out exploration with the probability 1-epsilon. We looked at the UCB algorithm, where we picked up the best action with the maximum upper bound value, followed by the TS algorithm, where we picked up the best action via beta distribution.

In the upcoming chapters, we will learn about deep learning and how deep learning is used to solve RL problems.