AI and the World's Hard Problems · Physics · arXiv · 2026-10-23
Graph-Recurrent Learning of Emergent Collective Dynamics
Collective ordering in correlated electron systems may not follow a closed, memoryless description. A graph-recurrent neural network combines gated recurrent memory with graph message passing and learns the dynamics directly from trajectories, without microscopic equations of motion. Applied to charge-density-wave order in the semiclassical Holstein model, it reproduces temporal and spatial correlations and the dynamical scaling during domain coarsening.