Atomic Descriptors: Symmetry and Locality

This lecture delves into the concept of atomic descriptors, focusing on the family of descriptors for specific body orders. The instructor explains the practical computation of correlations between atomic environments, emphasizing the importance of symmetry and locality in describing molecular structures. The lecture covers the construction of similarity kernels, the use of spherical harmonics as a basis for expanding properties, and the challenges of incorporating electrostatics in machine learning models for chemistry. The instructor showcases how the general framework of building correlation functions can be applied to various scientific domains, highlighting the synergy between physics principles and data-driven modeling.