Machine learning in atomistic simulations: from reaction pathways to phase diagrams

This lecture by the instructor covers the application of machine learning in atomistic simulations, focusing on topics such as modeling water, challenges of atomistic simulation, neural network potentials for water, and structure recognition using neural networks. The lecture also delves into the training process with Kalman filters, Gibbs free energy calculations, and the determination of local structures. Various water and ice models, as well as the impact of nuclear quantum effects on density, are discussed. The presentation concludes with insights on the melting/freezing of Au nanoparticles, local structure determination, and the effectiveness of bond order parameters in ice prediction.