Training sets based on uncertainty estimates in the cluster-expansion method

Abstract Cluster expansion (CE) has gained an increasing level of popularity in recent years, and its applications go far beyond original root binary alloys, reaching even complex crystalline systems often used energy materials research. Similar to other modern machine learning approaches science, many strategies have been proposed for training fitting the CE models first-principles calculation results. Here, we propose a new strategy constructing set based on their relevance Monte Carlo sampling statistical analysis reduction expected error. The model constructed from approach lower dependence specific details set, thereby reproducibility model. same method can be applied where it is desirable sample relevant configurational space with small data, which case when they consist calculations.