Bayesian optimization for goal-oriented multi-objective inverse material design

Abstract

Bayesian optimization (BO) can accelerate material design requiring timeconsuming experiments. However, although most material designs require tuning of multiple properties, the efficiency of multi-objective (MO) BO in timeconsuming experimental material design remains unclear, due to the complexity of handling multiple objectives. This study introduces MO BO method that efficiently achieves predefined goals and shows that by focusing on achieving the goals, BO can efficiently accelerate realistic MO design problems with small efforts. Benchmarks showed that the proposed BO method dramatically reduced the number of experiments needed to achieve goals relative to a baseline method. Virtual MO inverse design experiments with realistic material design problems were also performed, during which the proposed method could achieve goals within only around ten experiments in average and showed over 1000-fold acceleration relative to the random sampling for the most difficult case. The introduction of goal-oriented BO will precede real-world application of BO.

Rapid Discovery of Efficient Long-Wavelength Emission Garnet:Cr NIR Phosphors via Multi-Objective Optimization

“…The goal of MOO is to obtain a set of solutions that provide the best tradeoff between competing objectives . The utility function used in this work has two components: PAs and the weighted sum of the PAs . The PA is extended from the lower confidence bound (LCB) …”

“… LCB ( x ) = μ ( x ) − k σ ( x ) where μ( x ) is the mean of the predictions, σ( x ) is the standard deviation of the predictions, and k controls the strength of the exploration. The PA for a single objective can be calculated using eq : , PA ( x ) = 1 − normalΦ ( g − μ false( x false) σ false( x false) ) where μ( x ), σ( x ), and g are the mean of the predictions, the standard deviation of the predictions, and the desired predefined target value, respectively. Φ is the cumulative distribution function.…”

Materiomically Designed Polymeric Vehicles for Nucleic Acids: Quo Vadis?

“…251 In theory, Bayesian optimization is well suited for small initial data sets and experimental systems where this initial data can be expanded significantly to hundreds of candidates. 252 However, the potential of Bayesian optimization cannot be realized in polymeric gene delivery unless accompanied by advances in automating polymer synthesis, characterization, and biological testing. 4.5.…”

“…Bayesian optimization approaches are promising because they accommodate nonbinary responsesoffering a richer way of looking at the data beyond simple yes/no questionsand expert knowledge . In theory, Bayesian optimization is well suited for small initial data sets and experimental systems where this initial data can be expanded significantly to hundreds of candidates . However, the potential of Bayesian optimization cannot be realized in polymeric gene delivery unless accompanied by advances in automating polymer synthesis, characterization, and biological testing.…”

Bayesian vs Evolutionary Optimisation in Exploring Pareto Fronts for Materials Discovery

“…On top of these, there are also hybrid variants such as TSEMO [60] or MOEA/D-EGO [61] which integrate the use of MOEAs to improve the prediction quality of the underlying surrogate models. In general, BO as an overarching optimisation strategy has already been established as an attractive strategy for use in both computational design problems [62]- [67], as well as experimentation problems [68]- [74] due to its sample efficient approach.…”