3 Problem Statement

  • payoff function and auxiliary functions ("sim")
  • joint distr, multiple-output Gaussian process
  • reward: real, positive. sim, 0
  • policy, budget
  • objective: identify an optimal action

4 The Multi-fidelity BO Framework

  • 2 phases: explore and optimize
  • information gain
  • greedy, stopping conditions
  • optimize, update the posterior
  • choose the estimated optimum