Equivariant Neural Fields

Equivariant Neural Fields (ICLR 2025) — geometry-grounded conditional neural fields built on Bekkers' PONITA weight-sharing framework (PONITA itself ICLR 2024). The Equivariant Eikonal Networks follow-up (2025-05, preprint-stage) does grid-free seismic travel-time prediction that generalizes across velocity models. AMLab UvA (Wessels, Knigge, Vadgama, Gavves, Bekkers).