Open access
Aug 2026
Stability-Regularized Residual Neural ODEs: From Rollout-Error Contraction Diagnostics to a Train-Time Method for Robust Long-Horizon Forecasting
This work augments the one-step objective with a contraction penalty λEx[(μ2(Jf^(x))−c)+], and studies when this improves robust forecasting across stable, expansive, marginal, and chaotic regimes under realistic sensor-corruption noise.
Qin Li, Min Wan
· Mathematics · 0 citations