Which slot machine should you play at a casino to maximize your winnings? How can you develop a strategy to intelligently try out multiple slot machines and narrow down the most profitable machine? What does this problem have to do with BayesOpt? These are the questions this chapter will help us answer.
Chapter 4 was our introduction to BayesOpt policies, which decide how the search space should be explored and inspected. The exploration strategy of a BayesOpt policy should guide us toward the optimum of the objective function we’d like to optimize. The two particular policies we learned about were Probability of Improvement (PoI) and Expected Improvement (EI), which use the idea that we’d like to improve from the best objective value we have seen so far. This improvement-based mindset is only a heuristic and, therefore, doesn’t constitute the only approach to BayesOpt.