Using machine learning to process experimental data

August 21, 2019

Understanding the dynamical processes that govern the performance of functional materials is essential for the design of next generation materials to tackle global energy and environmental challenges. Many of these processes involve the dynamics of individual atoms or small molecules in condensed phases. Molecular dynamics simulations remain one of the few tools for probing dynamical processes; however, due to the large amounts of data generated in each simulation, searching for relevant dynamics is "much worse than searching for a needle in a haystack."

Professors Jeff Grossman and Yang Shao-Horn's paper in Nature Communications describes the development of graph dynamical networks, an unsupervised learning approach for understanding dynamics from molecular dynamics simulations. This method now automates the process of learning from the experimental data, making the task of searching for relevant and important data much easier, and potentially finding insights that were previously overlooked.