Navigating in High-Dimensional Search Space: A Hierarchical Bayesian Optimization Approach
This work addresses a known bottleneck in Bayesian optimization for high-dimensional problems, offering an incremental improvement with domain-specific applications in database tuning.
The paper tackles the challenge of optimizing black-box functions in high-dimensional search spaces by introducing HiBO, a hierarchical Bayesian optimization algorithm that integrates global-level search space partitioning into local acquisition strategies, resulting in outperforming state-of-the-art methods on synthetic benchmarks and showing practical effectiveness in tuning database management system configurations.
Optimizing black-box functions in high-dimensional search spaces has been known to be challenging for traditional Bayesian Optimization (BO). In this paper, we introduce HiBO, a novel hierarchical algorithm integrating global-level search space partitioning information into the acquisition strategy of a local BO-based optimizer. HiBO employs a search-tree-based global-level navigator to adaptively split the search space into partitions with different sampling potential. The local optimizer then utilizes this global-level information to guide its acquisition strategy towards most promising regions within the search space. A comprehensive set of evaluations demonstrates that HiBO outperforms state-of-the-art methods in high-dimensional synthetic benchmarks and presents significant practical effectiveness in the real-world task of tuning configurations of database management systems (DBMSs).