Newton-PINet:用 Newton 线性化加速 PINN 求解非线性 PDE
Newton-PINet: A fast physics-informed neural network with Newton linearization for meta-learning nonlinear PDEs
PINN 处理非线性 PDE 一直又慢又难训,这篇把它压到 5 分钟训练、误差降三个数量级,还开源了代码。做科学计算或仿真求解器的开发者可以直接试,比堆数据训练算子学习模型省得多。
研究人员提出 Newton-PINet,一种面向参数化非线性偏微分方程(PDE)元学习的物理信息神经网络框架,代码与数据集已在 GitHub 开源。
该方法包含三项核心设计:带 skip connections 的物理信息多层网络且输出层权重由最小二乘求解;两阶段训练策略,先用梯度训练学习跨任务的稳健表示,再对输出层做免梯度微调以快速泛化到新任务;
以及引入 Newton 线性化加速非线性问题的最小二乘迭代。在非线性 reaction-diffusion 基准上,Newton-PINet 的相对误差比近期神经求解器最多降低三个数量级,训练任务用量减少 16 倍,训练时间缩短一个数量级以上,约 5 分钟对比数小时。
相比依赖大规模数据集的算子学习方法(如数据驱动的 DeepONet/FNO 类方法),Newton-PINet 无需训练数据、直接以物理方程残差为约束,同时缓解了传统 PINN 优化难、泛化差的问题,对求解参数化 PDE 族的科研与工程团队是低成本的替代路径。
Newton-PINet: A fast physics-informed neural network with Newton linearization for meta-learning nonlinear PDEs
Scientific machine learning has opened new avenues for solving parameterized partial differential equations (PDEs), enabling models to learn a family of PDEs and generalize to unseen instances. In this context, data-driven operator learning methods typically require large training datasets, while physics-informed neural networks (PINNs) suffer from difficult optimization and limited generalization, especially for nonlinear PDEs. We propose Newton-PINet, a physics-informed network enhanced by Newton linearization, offering an effective meta-learning framework for nonlinear PDEs. Newton-PINet (i) employs a physics-informed multilayer network with skip connections, where the output-layer weights are solved by least squares; (ii) adopts a two-stage learning strategy that first leverages gradient-based training to learn robust representations from the available training tasks, and then performs gradient-free fine-tuning on the output layer for fast task-specific generalization; and (iii) incorporates a Newton linearization method to speed up the least-squares iteration for nonlinear PDE problems. On a challenging nonlinear reaction-diffusion benchmark, Newton-PINet achieves up to three orders of magnitude lower relative error than recent neural solvers, while using 16x fewer training tasks and over an order of magnitude less training time (about 5 minutes versus several hours). This work advances the meta-learning of PINNs toward data-efficient, fast, and generalizable physics solvers. The code and datasets are publicly available at: https://github.com/fycAIwj/Newton-PINet.
This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include PINN, PDE solving, meta-learning, open-source models.
Source attribution: curated from arXiv cs.AI (@Yuchen Fan, Chang Wei, Pao-Hsiung Chiu, Chin Chun Ooi, Heyang Wang, Jian Cheng Wong).