{"id":"arxiv/2608.13132","url":"https://arxiv.org/abs/2608.13132","tags":["math.NA"],"score":0,"title":"Analysis of Error Propagation in Autoencoder-Based Reduced-Order Neural Ordinary Differential Equations","domain":"math.NA","source":"arxiv","authors":["Jingyi Zhang","Gwanghyun Jo"],"reviews":0,"subject":"math.NA","summary":"Neural ODE reduced-order models often achieve comparable local prediction accuracy, yet their long-horizon extrapolation behavior can differ substantially. To analyze this discrepancy, we develop a path-integral identity that separates local discrepancy injection from amplification in the learned latent dynamics. The associated multi-step Jacobian norms quantify transport sensitivity and distinguish different propagation regimes. Experiments on the Burgers and Gray--Scott systems exhibit two distinct patterns of error evolution. In Burgers systems, prediction errors remain bounded and are…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-13","reactions":{}}