Classical numerical optimization assumes that the functions under consideration are easily accessible and inexpensive to evaluate.
However, some real-world functions can take months and more to evaluate (take the efficiency of a medication as example) such that standard optimizers would be too lavish with function evaluations.
Bayesian optimizers aim to reduce the number of function evaluations by making use of prior information and being less concerned with computational efficiency.
This talk is meant as a small introduction into (discrete) Bayesian optimization, presenting basic building blocks and tools, and resources where to find more information.
