Relaxing the Additivity Constraints in Decentralized No-Regret High-Dimensional Bayesian Optimization

TL;DR

This paper introduces DuMBO, a decentralized Bayesian Optimization algorithm that relaxes additive structure assumptions and mitigates over-exploration, enabling efficient high-dimensional optimization without sacrificing theoretical guarantees.

Contribution

It relaxes additive assumptions in high-dimensional BO and addresses over-exploration, providing an asymptotically optimal decentralized algorithm.

Findings

DuMBO performs competitively with state-of-the-art BO algorithms.

It effectively handles high-dimensional additive structures.

The method maintains theoretical guarantees without restrictive assumptions.

Abstract

Bayesian Optimization (BO) is typically used to optimize an unknown function $f$ that is noisy and costly to evaluate, by exploiting an acquisition function that must be maximized at each optimization step. Even if provably asymptotically optimal BO algorithms are efficient at optimizing low-dimensional functions, scaling them to high-dimensional spaces remains an open problem, often tackled by assuming an additive structure for $f$. By doing so, BO algorithms typically introduce additional restrictive assumptions on the additive structure that reduce their applicability domain. This paper contains two main contributions: (i) we relax the restrictive assumptions on the additive structure of $f$ without weakening the maximization guarantees of the acquisition function, and (ii) we address the over-exploration problem for decentralized BO algorithms. To these ends, we propose DuMBO, an asymptotically optimal decentralized BO algorithm that achieves very competitive performance against state-of-the-art BO algorithms, especially when the additive structure of $f$ comprises high-dimensional factors.