Update of PHYSBO: Improving Usability and Portability of Bayesian Optimization for Physics and Materials Research

TL;DR

This paper presents the major updates introduced in PHYSBO versions 2 and 3, with a focus on improvements in usability, portability, and practical deployment rather than on new optimization algorithms.

Abstract

Bayesian optimization (BO) is widely used to accelerate physics and materials research, where objective function evaluations are computationally or experimentally expensive. While many BO frameworks focus on algorithmic efficiency, practical usability and portability are equally critical for sustained use in real research environments. PHYSBO is a Bayesian optimization library designed to address these needs by enabling optimization over user-defined candidate pools and by supporting domain-specific problem settings. This paper presents the major updates introduced in PHYSBO versions 2 and 3, with a focus on improvements in usability, portability, and practical deployment rather than on new optimization algorithms. In PHYSBO version 2, the software license was changed from GPL to MPL to improve compatibility with a wider range of research and software ecosystems. Building on this revision, PHYSBO version 3 introduces a set of implementation-oriented updates aimed at improving usability and portability, without modifying the core optimization algorithms. These updates include improvements in computational performance and scalability, extended support for multi-objective optimization, the introduction of range-based policies for continuous-variable optimization, the removal of environment-dependent components such as tightly coupled Cython modules, and compatibility with NumPy 2. These improvements reduce the technical and organizational burden on users, enabling PHYSBO to be deployed across diverse computing environments and research workflows. By emphasizing portability and ease of integration while maintaining sufficient performance, PHYSBO version 3 is positioned as a sustainable research infrastructure for Bayesian optimization in physics and materials science.

Figures (6)

• Figure 1: Objective function landscapes of the VLMOP2 benchmark. Left: $-y_1(\boldsymbol{x})$. Right: $-y_2(\boldsymbol{x})$. The two objectives attain their maxima at different locations in the design space, inducing a trade-off between them.
• Figure 2: Pareto front comparison for the VLMOP2 benchmark. Blue points denote evaluated solutions, while orange markers indicate the non-dominated solutions. HVPI yields the most complete coverage of the Pareto front, NDS provides a competitive approximation at a significantly lower computational cost, and ParEGO produces a sparse set of trade-off solutions.
• Figure 3: Objective function landscapes of the Kita--Yabumoto--Mori--Nishikawa (KYMN) benchmark. Left: $y_1(x_1, x_2)$. Right: $y_2(x_1, x_2)$. The feasible region is implicitly restricted by nonlinear inequality constraints, leading to a highly constrained optimization landscape.
• Figure 4: Pareto front comparison for the Kita--Yabumoto--Mori--Nishikawa benchmark. Blue points denote evaluated solutions, and orange markers indicate the non-dominated solutions. ParEGO concentrates sampling on a narrow feasible Pareto-optimal region, whereas NDS yields a broader distribution of near-Pareto solutions. HVPI exhibits higher computational cost and reduced coverage in this constrained scenario.
• Figure 5: Dependence of the computational time for 40 Bayesian optimization steps on the number of objective functions $p$ for each method. The mean and standard deviation over 10 independent runs are shown.