[论文解读] SGNO: Spectral Generator Neural Operators for Stable Long Horizon PDE Rollouts

Jiayi Li, Zhaonan Wang+1 more|arXiv (Cornell University)|Feb 21, 2026

Model Reduction and Neural Networks被引用 0

一句话总结

SGNO 引入一种光谱、指数积分器神经算子，具有非正实部生成器和门控 forcing，以稳定长时域PDE滚动预测，在七个 APEBench 任务上实现最先进的滚动精度。

ABSTRACT

Neural operators provide fast PDE surrogates and often generalize across parameters and resolutions. However, in the short train long test setting, autoregressive rollouts can become unstable. This typically happens for two reasons: one step errors accumulate over time, and high frequency components feed back and grow. We introduce the Spectral Generator Neural Operator (SGNO), a residual time stepper that targets both effects. For the linear part, SGNO uses an exponential time differencing update in Fourier space with a learned diagonal generator. We constrain the real part of this generator to be nonpositive, so iterating the step does not amplify the linear dynamics. For nonlinear dynamics, SGNO adds a gated forcing term with channel mixing within each Fourier mode, which keeps the nonlinear update controlled. To further limit high frequency feedback, SGNO applies spectral truncation and an optional smooth mask on the forcing pathway. We derive a one step amplification bound and a finite horizon rollout error bound. The bound separates generator approximation error from nonlinear mismatch and gives sufficient conditions under which the latent $L^2$ norm does not grow across rollout steps. On APEBench spanning 1D, 2D, and 3D PDE families, SGNO achieves lower long horizon error and longer stable rollout lengths than strong neural operator baselines. Ablations confirm the roles of the generator constraint, gating, and filtering.The code is available at https://github.com/lijy32123-cloud/SGNO.

研究动机与目标

• 在短期训练数据下，为时依赖PDE替代模型提供鲁棒的长时域预测动机。
• 开发一个显式控制线性放大和高频反馈的神经算子。
• 提出一个基于光谱 ETD 的时间步进器，具备稳定性保证和实用的截断/遮罩。
• 在 APEBench 的线性、非线性及反应扩散PDEs 上对 SGNO 进行与强基线的实证比较。

提出的方法

• 在潜在空间中以光谱 ETD 时间推进块描述半线性PDE动力学。
• 在傅里叶空间使用具有非正实部的学习对角生成器，以稳定重复步。
• 通过 φ_1 加权 forcing 项以及跨傅里叶模态的门控和通道混合来引入非线性效应。
• 对 forcing 通路应用光谱截断和可选的平滑掩蔽以抑制高频反馈。
• 给出一步放大界限和有限时域滚动误差递归，将架构与稳定性联系起来。
• 采用一步教师强制进行训练，并在固定分辨率和时间步下评估长时滚动。

Figure 1 : (a) Overall architecture of SGNO. The input state is lifted to a latent feature space, propagated through stacked time advance blocks, and projected back to obtain the next state. (b) Time advance block. Features are propagated by a stabilized spectral ETD operator in the Fourier domain,

实验结果

研究问题

• RQ1 一种带有光谱稳定 ETD 更新的神经算子，是否能在不同PDE动力学下实现稳定的长时滚动？
• RQ2 哪些架构组件（生成器约束、门控、过滤）对抑制误差放大和高频反馈至关重要？
• RQ3 光谱截断和掩蔽是否显著提升自回归PDE预测在长时间尺度上的稳定性和准确度？
• RQ4 SGNO 的稳定性界限与线性、非线性及反应扩散PDEs 的实际长滚动性能之间有什么关系？

主要发现

• SGNO 在七个任务中有六个任务取得最佳的 GMean100 滚动指标，在 3D Swift–Hohenberg 任务中位列第二。
• 消融研究表明 forcing 注入（α_g）对长时稳定性尤为关键，点对点修正（α_w）提供额外益处。
• SGNO 能显著延迟长滚动中的不稳定性起始（如在 1DKdV 中提高了稳定步分布）。
• 光谱生成器约束确保迭代下线性动力学的不放缩性，降低每步放大。
• 光谱截断与掩蔽选项进一步减轻高频反馈，同时不牺牲准确性。
• 与 Fourier Neural Operator 基线相比，SGNO 的参数量相当且每步推理开销适中。

Figure 2 : Long-horizon stability on 1D KdV. Top: nRMSE trajectories (first 100 steps shown for readability); both baseline and SGNO show the per-step median across test trajectories with a $p10$ – $p90$ band. Bottom: empirical CDF of per-trajectory stable steps computed on full $T_{\mathrm{eval}}=2