Gia-Wei Chern

Gauge-equivariant GNNs generate internal representations that exhibit a surprising loop geometry, with contour length growing exponentially while spatial extent grows only logarithmically with depth.

GNNs can revolutionize spin dynamics simulations by learning effective magnetic force fields that match electronic calculations with unprecedented efficiency.

Gauge-equivariant GNNs unlock the ability to learn intrinsically nonlocal observables in lattice gauge theories by directly embedding non-Abelian symmetries into message passing.

Machine learning can now accurately model complex magnetization dynamics in metallic spin systems, even capturing non-equilibrium effects and paving the way for quantum-accurate spintronics simulations.

GNNs offer a surprisingly simple and scalable way to build symmetry-aware force fields for simulating the dynamics of correlated lattice systems, outperforming descriptor-based neural networks.