Loss Landscape Geometry of Partial Differential Equation Emulators: Or, Symmetry Learning via Gradient Alignment

arXiv:2601.20172v2 Announce Type: replace Abstract: We study how neural emulators of partial differential equation solution operators internalize physical symmetries by introducing an influence-based diagnostic that measures the propagation of parameter updates between symmetry-related states, defined as the metric-weighted overlap of loss gradients evaluated along group orbits. This quantity probes the local geometry of the learned loss landscape and goes beyond forward-pass equivariance tests by directly assessing whether learning dynamics couple physically equivalent configurations. Applying our diagnostic to autoregressive fluid flow emulators, we show that orbit-wise gradient coherence provides the mechanism for learning to generalize over symmetry transformations and indicates when training selects a symmetry compatible basin. The result is a novel technique for evaluating if surrogate models have internalized symmetry properties of the known solution operator.

Related

• Weak Correlations as the Underlying Principle for Linearization of Gradient-Based Learning Systems
• Discovering Physical Directions in Weight Space: Composing Neural PDE Experts
• MiNO: Cotangent-bundle propagator learning for PDEs

Source: arXiv cs.LG | 2026-08-27