Accepted · September 24, 2026
Dongzhe Zheng, Christine Allen-Blanchette
NeurIPS 2026 Spotlight (Top 1.3% of submissions; Top 5% of accepted papers)
Riemannian Hodge Message Passing (RHMP) learns positive-definite metrics on discrete field spaces, making geometry adaptable while keeping topological identities exact, Hodge energies nonnegative, and propagation equivariant under orthogonal changes of hidden feature basis.
Accepted · ICML 2026
Dongzhe Zheng, Tao Zhong, Christine Allen-Blanchette
International Conference on Machine Learning (ICML) 2026 · Regular Paper
Introducing a hybrid Eulerian–Lagrangian architecture, Hodge Spectral Duality (HSD), that exploits Hodge orthogonality to decouple unlearnable topological features from learnable geometric dynamics. Combining discrete differential forms with an auxiliary ambient space, the method markedly improves accuracy, efficiency, and fidelity to physical invariants when solving field equations on geometric meshes.