Robust Entropy Search for Safe Efficient Bayesian Optimization

TL;DR

This paper introduces Robust Entropy Search (RES), a novel acquisition function for Bayesian Optimization that enhances robustness against adversarial perturbations, leading to more reliable and efficient optimization in engineering applications.

Contribution

The paper proposes RES, an information-based acquisition function that improves robustness in Bayesian Optimization, addressing adversarial perturbations and outperforming existing methods.

Findings

RES reliably finds robust optima in synthetic and real data.

RES outperforms state-of-the-art algorithms in robustness and efficiency.

Empirical results demonstrate the effectiveness of RES in practical scenarios.

Abstract

The practical use of Bayesian Optimization (BO) in engineering applications imposes special requirements: high sampling efficiency on the one hand and finding a robust solution on the other hand. We address the case of adversarial robustness, where all parameters are controllable during the optimization process, but a subset of them is uncontrollable or even adversely perturbed at the time of application. To this end, we develop an efficient information-based acquisition function that we call Robust Entropy Search (RES). We empirically demonstrate its benefits in experiments on synthetic and real-life data. The results showthat RES reliably finds robust optima, outperforming state-of-the-art algorithms.