Bayesian Optimization on Function Spaces via Sparse RKHS Manifolds

Bayesian Optimization (BO) has become an established methodology for minimizing black-box functions of a vector input. Often, however, this parameter vector arises from the discretization of an inherently functional relationship.