CAGES: Cost-Aware Gradient Entropy Search for Efficient Local Multi-Fidelity Bayesian Optimization

arXiv:2405.07760v1 Announce Type: new
Abstract: Bayesian optimization (BO) is a popular approach for optimizing expensive-to-evaluate black-box objective functions. An important challenge in BO is its application to high-dimensional search spaces due in large part to the curse of dimensionality. One way to overcome this challenge is to focus on local BO methods that aim to efficiently learn gradients, which have shown strong empirical performance on a variety of high-dimensional problems including policy search in reinforcement learning (RL). However, current local …

abstract application arxiv bayesian box challenge cost cs.lg dimensionality entropy fidelity focus functions gradient optimization part popular search spaces stat.ml the curse of dimensionality type