<ul><li>Researchers introduce Equivariant Neural Eikonal Solvers, a framework integrating Equivariant Neural Fields with Neural Eikonal Solvers for scalable travel-time prediction on homogeneous spaces.</li><li>The approach uses a single neural field conditioned on signal-specific latent variables represented as point clouds in a Lie group to model diverse Eikonal solutions.</li><li>Integration of Equivariant Neural Fields ensures equivariant mapping from latent representations to the solution field, providing enhanced representation efficiency, robust geometric grounding, and solution steerability.</li><li>The framework, coupled with Physics-Informed Neural Networks, accurately models Eikonal travel-time solutions while generalizing to arbitrary Riemannian manifolds with regular group actions, demonstrating superior performance in seismic travel-time modeling.</li></ul>

Equivariant Eikonal Neural Networks: Grid-Free, Scalable Travel-Time Prediction on Homogeneous Spaces

Discover more