BONSAI: Bayesian Optimization with Natural Simplicity and Interpretability
This addresses a practical issue for practitioners tuning parameters in applications where defaults are carefully engineered, though it is incremental as it builds on standard BO methods.
The paper tackles the problem of Bayesian optimization (BO) often deviating unnecessarily from default configurations, making recommendations hard to vet, by introducing BONSAI, a default-aware BO policy that prunes low-impact deviations; empirically, BONSAI substantially reduces non-default parameters while maintaining competitive optimization performance.
Bayesian optimization (BO) is a popular technique for sample-efficient optimization of black-box functions. In many applications, the parameters being tuned come with a carefully engineered default configuration, and practitioners only want to deviate from this default when necessary. Standard BO, however, does not aim to minimize deviation from the default and, in practice, often pushes weakly relevant parameters to the boundary of the search space. This makes it difficult to distinguish between important and spurious changes and increases the burden of vetting recommendations when the optimization objective omits relevant operational considerations. We introduce BONSAI, a default-aware BO policy that prunes low-impact deviations from a default configuration while explicitly controlling the loss in acquisition value. BONSAI is compatible with a variety of acquisition functions, including expected improvement and upper confidence bound (GP-UCB). We theoretically bound the regret incurred by BONSAI, showing that, under certain conditions, it enjoys the same no-regret property as vanilla GP-UCB. Across many real-world applications, we empirically find that BONSAI substantially reduces the number of non-default parameters in recommended configurations while maintaining competitive optimization performance, with little effect on wall time.