A multi-fidelity machine learning approach to high throughput materials screening

Abstract

The ever-increasing capability of computational methods has resulted in their general acceptance as a key part of the materials design process. Traditionally this has been achieved using a so-called computational funnel, where increasingly accurate - and expensive – methodologies are used to winnow down a large initial library to a size which can be tackled by experiment. In this paper we present an alternative approach, using a multi-output Gaussian process to fuse the information gained from both experimental and computational methods into a single, dynamically evolving design. Common challenges with computational funnels, such as mis-ordering methods, and the inclusion of non-informative steps are avoided by learning the relationships between methods on the fly. We show this approach reduces overall optimisation cost on average by around a factor of three compared to other commonly used approaches, through evaluation on three challenging materials design problems.

Multi-fidelity Bayesian optimization of covalent organic frameworks for xenon/krypton separations

“…MFBO (or its parent, multi-informationsource BO 73 ) has been scarcely applied to materials discovery. 72,[74][75][76][77]…”

Multi-fidelity machine learning for predicting bandgaps of nonlinear optical crystals

“…Second, the selection of different M 2 candidates in step 4(ii) can be further optimized, for example, by using Bayesian approach. 28 Finally, although we limited our scope to the bandgap of NLO crystals in this work, the benchmark using other functional crystals and other material properties is necessary to validate the generalization of the present multi-fidelity ML approach.…”

“…27 Fare et al presented a multi-fidelity Bayesian model, in which the relationship between different fidelities can be captured on the fly, and the hierarchy of methods at different fidelities was no longer necessary in priority. 28 Very recently, Jacobs et al investigated how to maximize the multi-fidelity data advantage and optimized the data acquisition ratio for materials discovery. 29 Remarkably, not only the predictive performance at each fidelity but also the interplay between different fidelities acts as a key to multi-fidelity ML.…”

Atlas: A Brain for Self-driving Laboratories

“…Multi-fidelity strategies are quickly becoming a popular approach for resource-intensive problems in chemistry and materials science. [113][114][115][116][117][118][119] Atlas provides a MultiFidelityPlanner based on the trace-aware knowledge gradient 120,121 and augmented-EI (aEI) acquisition functions 122 which allows for the inclusion of an arbitrary number of information sources with discrete fidelity levels.…”