In the previous chapter, we introduced GPyTorch, the backbone horsepower used by BoTorch to obtain the GP posterior using PyTorch. This paves the way for our introduction to BoTorch, the main topic in this chapter. Specifically, we will focus on how it implements the expected improvement acquisition function covered in Chapter 3 and performs the inner optimization in search of the next best proposal for sampling location.
In addition, we will also cover another type of acquisition ...