Asynchronous Batch Bayesian Optimization with Pipelining Evaluations for Experimental Resource$\unicode{x2013}$constrained Conditions

TL;DR

This work tackles the inefficiency of traditional batch Bayesian optimization in settings with limited parallel hardware by introducing PipeBO, a pipelined, asynchronous Bayesian optimization framework. PipeBO partitions each experiment into K sequential processes and allows P parallel experiments, updating running process parameters with results from completed experiments via a local penalizer and a GP-UCB–based acquisition. The approach is formalized with a clear problem setting and algorithm, and demonstrated on 24 BBOB benchmark functions, showing substantial reductions in elapsed optimization time (e.g., ~56% at K=2, ~50% at K=3, ~38% at K=5). The findings indicate PipeBO can enable faster, data-driven optimization in resource-constrained laboratories and can potentially be enhanced further with scheduling techniques for heterogeneous process times.

Abstract

Bayesian optimization is efficient even with a small amount of data and is used in engineering and in science, including biology and chemistry. In Bayesian optimization, a parameterized model with an uncertainty is fitted to explain the experimental data, and then the model suggests parameters that would most likely improve the results. Batch Bayesian optimization reduces the processing time of optimization by parallelizing experiments. However, batch Bayesian optimization cannot be applied if the number of parallelized experiments is limited by the cost or scarcity of equipment; in such cases, sequential methods require an unrealistic amount of time. In this study, we developed pipelining Bayesian optimization (PipeBO) to reduce the processing time of optimization even with a limited number of parallel experiments. PipeBO was inspired by the pipelining of central processing unit architecture, which divides computational tasks into multiple processes. PipeBO was designed to achieve experiment parallelization by overlapping various processes of the experiments. PipeBO uses the results of completed experiments to update the parameters of running parallelized experiments. Using the Black-Box Optimization Benchmarking, which consists of 24 benchmark functions, we compared PipeBO with the sequential Bayesian optimization methods. PipeBO reduced the average processing time of optimization to about 56% for the experiments that consisted of two processes or even less for those with more processes for 20 out of the 24 functions. Overall, PipeBO parallelizes Bayesian optimization in the resource-constrained settings so that efficient optimization can be achieved.

Figures (8)

• Figure 1: Conceptual diagram of parallelization through pipelining
• Figure 1: Problem sets tested in this study
• Figure 2: Updating parameters during the experiment
• Figure 2: Optimization process for 24 benchmark functions ($K=2$)
• Figure 3: Comparison between Bayesian optimization (BO) without pipelining and PipeBO