Practical Path-based Bayesian Optimization

TL;DR

This paper extends Bayesian optimization to efficiently handle both experimental and input change costs, including constraints and multi-objective scenarios, improving experimental design in chemical and drug manufacturing.

Contribution

It introduces an extension of the SnAKe algorithm that manages dual cost types and incorporates input change constraints and multi-objective optimization.

Findings

Effective handling of experimental and input change costs.

Enhanced optimization under input change constraints.

Applicability to multi-objective experimental design.

Abstract

There has been a surge in interest in data-driven experimental design with applications to chemical engineering and drug manufacturing. Bayesian optimization (BO) has proven to be adaptable to such cases, since we can model the reactions of interest as expensive black-box functions. Sometimes, the cost of this black-box functions can be separated into two parts: (a) the cost of the experiment itself, and (b) the cost of changing the input parameters. In this short paper, we extend the SnAKe algorithm to deal with both types of costs simultaneously. We further propose extensions to the case of a maximum allowable input change, as well as to the multi-objective setting.