analiza aiqbench
This research introduces RHMP, a novel neural architecture that strictly separates topological conservation laws from learned geometric metrics in physical field simulations on meshes. By enforcing cochain-frame equivariance and exact complex identities, the method outperforms existing surrogates across diverse benchmarks including fluids, electromagnetism, and CFD without mixing these distinct roles. The work represents significant progress in machine learning for scientific computing by ensuring that data-driven models respect fundamental mathematical structures inherent to physical systems.
Learning Discrete Riemannian Metrics for Physical Fields with Cochain-Frame Equivarianc
arXiv:2608.14556v1 Announce Type: new Abstract: Physical fields on meshes require a separation between topology and geometry: conservation laws are topological and should be exact, while geometry, material response, and anisotropic coupling must be learned from data. Existing neural surrogates often mix these roles inside unconstrained message passing. We introduce Riemannian Hodge Message Passing (RHMP), which turns this separation into an architectural principle. RHMP fixes the cellular coboundaries ($d_k$) determined by oriented incidence and learns symmetric positive-definite cochain metrics ($H_k$) for geometry-dependent propagation. Treating $H_k$ as the learned metric motivates cochain-frame equivariance: physical propagation should be invariant to orthogonal changes of the hidden cochain feature basis. RHMP implements this principle with metric-weighted Hodge blocks ($d_k^\top H_{k+1}d_k$), yielding exact cochain-complex identities ($d_{k+1}d_k=0$), nonnegative Hodge energies, positive-semidefinite operators, and exact Abelian curvature invariance. Across seven physical benchmarks spanning fluids, electromagnetism, gauge fields, and variable-mesh CFD, RHMP achieves the best overall performance, with the largest gains when topology, learned geometry, and field structure interact.
Abstract:Physical fields on meshes require a separation between topology and geometry: conservation laws are topological and should be exact, while geometry, material response, and anisotropic coupling must be learned from data. Existing neural surrogates often mix these roles inside unconstrained message passing. We introduce Riemannian Hodge Message Passing (RHMP), which turns this separation into an architectural principle. RHMP fixes the cellular coboundaries ($d_k$) determined by oriented incidence and learns symmetric positive-definite cochain metrics ($H_k$) for geometry-dependent propagation. Treating $H_k$ as the learned metric motivates cochain-frame equivariance: physical propagation should be invariant to orthogonal changes of the hidden cochain feature basis. RHMP implements this principle with metric-weighted Hodge blocks ($d_k^\top H_{k+1}d_k$), yielding exact cochain-complex identities ($d_{k+1}d_k=0$), nonnegative Hodge energies, positive-semidefinite operators, and exact Abelian curvature invariance. Across seven physical benchmarks spanning fluids, electromagnetism, gauge fields, and variable-mesh CFD, RHMP achieves the best overall performance, with the largest gains when topology, learned geometry, and field structure interact.
Machine Learning (cs.LG)
arXiv:2608.14556 [cs.LG]
(or arXiv:2608.14556v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.14556 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 247 cuvinte
Subiecte conexe
Quantum Simulation on Riemannian Manifolds
arXiv · 5 oct. 2026
Riemann Manifold Diffusion Simplifies Path Integral Construction
Quantum Zeitgeist · 21 aug. 2026
Physics-Informed Graph Neural Networks for Surface Code Decoding via Discrete Exterior Calculus
arXiv · 23 iul. 2026
GA-VINO: A Geometry-Aware Variational Physics-informed Neural Operator for Mindlin-Reissner Plates
arXiv · 20 iul. 2026