B3O: Scalable Boltzmann Batch Bayesian Optimization
2026-06-29 • Machine Learning
Machine Learning
AI summaryⓘ
The authors address the challenge of running many simulations at once to improve designs using a method called Bayesian Optimization (BO). They introduce B3O, a new way to pick multiple solutions by treating the selection process like sampling from a special probability distribution, avoiding slow or less diverse methods. They show theoretically that this approach works well without much downside and demonstrate it performs better than others in tests, including real-world problems like designing electrodes and configuring race cars. Their approach helps make large-scale, parallel optimization more efficient and reliable.
Bayesian Optimizationbatch optimizationBoltzmann distributionacquisition functionparallel simulationregret analysismulti-objective optimizationmixed-variable optimizationsampling methods
Authors
Maximilian Bloor, Liyuan Xu, Hrvoje Stojic, Victor Picheny
Abstract
Modern engineering workflows increasingly rely on massive parallel simulation, driving the need for scalable, large-batch Bayesian Optimization (BO). Existing batch BO methods, however, incur large computational cost or rely on approximations that erode batch diversity. We propose B3O (Boltzmann Batch Bayesian Optimization), a framework that reframes batch generation as a pure sampling problem: drawing samples directly from the Boltzmann distribution defined by the acquisition function avoids the bottlenecks of existing large-batch methods. Theoretically, we prove that queries sampled from this distribution incur only negligible additional regret. Empirically, B3O outperforms existing batch BO methods on standard synthetic benchmarks and adapts robustly across complex applied tasks, including multi-objective electrode design and mixed-variable race car configuration.