Mechanical Field Networks: Structured Neural Dynamics for Multivariate Systems

In three linesMF-Net is a recurrent dynamical model representing multivariate systems via a shared field state, updated by a learned relation law. On Lorenz-96 (40D), it achieves R²=0.798±0.018 at 8-step forecasting and recovers local coupling structure with perfect precision (1.000±0.000). The model combines forecasting with structural interpretability.

Summary generated by Claude — human-verified