Atomistic Machine Learning: Physics and Data

This lecture by the instructor covers the topic of Atomistic Machine Learning, emphasizing the integration of physical principles into machine learning models. It explores the use of symmetry-adapted representations, SOAP kernels, and the application of machine learning to predict molecular properties with high accuracy. The lecture delves into the challenges of learning tensorial properties, optimizing representations, and understanding the range of interactions in molecular systems. It also discusses the development of a data-driven periodic table of elements and the implications of machine learning in predicting molecular polarizabilities and non-covalent interactions.