Machine learning: Physics and Data

This lecture explores the intersection between physics and data in the context of machine learning, covering topics such as the thin line between physics and data, atomic cluster expansion force fields, automated identification of collective variables, and unsupervised learning in atomistic simulations. It delves into the challenges and advancements in fitting machine learning potentials, pioneers of fitting machine learning, and the importance of representation in machine learning models. The lecture also discusses regression, loss, and complexity in machine learning models, learning curves, and the significance of symmetry and locality in atomistic machine learning.