Bayesian Optimization for Materials Design with Mixed Quantitative and Qualitative Variables

Abstract

Although Bayesian Optimization (BO) has been employed for accelerating materials design in computational materials engineering, existing works are restricted to problems with quantitative variables. However, real designs of materials systems involve both qualitative and quantitative design variables representing material compositions, microstructure morphology, and processing conditions. For mixed-variable problems, existing Bayesian Optimization (BO) approaches represent qualitative factors by dummy variables first and then fit a standard Gaussian process (GP) model with numerical variables as the surrogate model. This approach is restrictive theoretically and fails to capture complex correlations between qualitative levels. We present in this paper the integration of a novel latent-variable (LV) approach for mixed-variable GP modeling with the BO framework for materials design. LVGP is a fundamentally different approach that maps qualitative design variables to underlying numerical LV in GP, which has strong physical justification. It provides flexible parameterization and representation of qualitative factors and shows superior modeling accuracy compared to the existing methods. We demonstrate our approach through testing with numerical examples and materials design examples. The chosen materials design examples represent two different scenarios, one on concurrent materials selection and microstructure optimization for optimizing the light absorption of a quasi-random solar cell, and another on combinatorial search of material constitutes for optimal Hybrid Organic-Inorganic Perovskite (HOIP) design. It is found that in all test examples the mapped LVs provide intuitive visualization and substantial insight into the nature and effects of the qualitative factors. Though materials designs are used as examples, the method presented is generic and can be utilized for other mixed variable design optimization problems that involve expensive physics-based simulations.

Machine-Learning-Assisted De Novo Design of Organic Molecules and Polymers: Opportunities and Challenges

“…Yet, the design of new molecules and materials involve both continuous/discretized and qualitative/quantitative design variables, representing molecular constituents, material compositions, microstructure morphology, and processing conditions. For these mixed variable design optimization problems, the existing BO approaches are usually restrictive theoretically and fail to capture complex correlations between input variable and output properties [207,208,232]. Therefore, new RL or BO methods should be formulated and developed to resolve these issues.…”

“…Lastly, designing the reward functions is difficult when many objectives are presented, i.e., multiple properties are to be optimized. For multi-variable optimization problems, Bayesian optimization is a favorable method [207,208].…”

“…Among the many global optimization methods reported in literature, BO stands out due to its capability of locating the global optima for highly non-linear functions within tens of objective function (i.e., material properties) evaluations. BO accomplishes this by repeating these three steps [207,208]:…”

Robotic Search for Optimal Cell Culture in Regenerative Medicine

“…However, many other cell types and differentiation targets lack established protocols, some of which probably demand a more sophisticated optimization technique able to deal with categorical values and their combinations. Such technique could optimize the structure of the protocol simultaneously with continuous parameter values while minimizing the execution costs incurred by the large number of possible combinations ( 21 ).…”

Bayesian task embedding for few-shot Bayesian optimization

“…A recent work by Zhang et al [31] also studies Bayesian optimization in the context of qualitative and quantitative inputs. However, they learn point embeddings via maximum likelihood estimation rather than inferring a posterior through Bayesian inference; this makes the overall model prone to overfitting in practice and generally unsuitable for providing credible uncertainty estimates.…”