Exascale computing is creating opportunities for simulations of chemical behavior that include the chemical realism of complex mixtures, non-ideality and extreme conditions. Physical and predictive models are increasingly benefitting from features extracted from simulation data that encode high-dimensional information in a lower-dimensional representation. This emerging and active area of research has immense opportunity for creating holistic information content about a chemical system that includes multiscale spatiotemporal correlation and relationships with the underlying energy landscapes that govern state distributions and transformations. Yet significant challenges remain for maintaining interpretability, understanding uncertainty and developing adaptive approaches for feature selection under varying phase space.
Copyright. All Rights Reserved.