Machine Learning for Voluminous Quantum Materials Data

Decades of efforts by the quantum materials research community drove a "data revolution." Modern experimental modalities produce high-dimensional data in large volumes. Unprecedented control and new facilities imply new dimension and new knobs, such as time-resolved probing or scanning probing. Moreover, massive amounts of high-throughput ab-initio data and curated experimental data are becoming accessible to researchers. Much needed are data-centric approaches that accelerate discoveries from these data through synergetic interaction with expert human researchers' insights. A synergy between data science and quantum materials research is essential for such endeavors to result in scientific progress. I will present cases of fruitful collaborations that led to new insights and started to shape an approach to data sets of the new era. Specifically, I will discuss how to use unsupervised learning to discover new physics from large volumes of evolving data and how to use supervised learning to uncover descriptors of emergent properties from limited volume of expertly curated data.