Lecture 23 - Bayesian Global Optimization

Objectives

• Recognize the challenges of optimizing expensive, potentially noisy, objective functions.

• Explain the concept of the value of information.

• Recognize the Bayesian global optimization algorithm.

• Develop intuition about some acquisition functions, such as the expected improvement, the probability of improvement, and the maximum upper interval.

• Quantify the epistemic uncertainty in the solution of an optimization problem.