arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.16845cs.LGcs.AIcs.NAmath.NA
收藏

具有局部积分与微分核的神经算子

Neural Operators with Localized Integral and Differential Kernels

California Institute of Technology(加州理工学院);NVIDIA(英伟达公司)

Miguel Liu-Schiaffini, Julius Berner, Boris Bonev, Thorsten Kurth, Kamyar Azizzadenesheli, Anima Anandkumar

AI总结:

本文提出一种结合局部微分与积分核的神经算子学习方法,通过改进FNO以捕捉局部特征,在湍流Navier-Stokes和球面浅水方程上将相对L2误差降低34-72%。

摘要:

神经算子学习函数空间之间的映射,这对于学习偏微分方程的解算子以及其他科学建模应用具有实际意义。其中,傅里叶神经算子(FNO)是一种流行的架构,它在傅里叶空间中执行全局卷积。然而,这种全局操作往往容易导致过度平滑,并且可能无法捕捉局部细节。相比之下,卷积神经网络(CNN)能够捕捉局部特征,但仅限于在单一分辨率下进行训练和推理。在这项工作中,我们提出了一种有原则的算子学习方法,该方法通过在学习微分算子和具有局部支撑核的积分算子的两种框架下捕捉局部特征。具体而言,受模板法(stencil methods)的启发,我们证明了在卷积神经网络核值的适当缩放条件下可以得到微分算子。为了获得局部积分算子,我们利用基于离散-连续卷积的合适基表示来处理核。这两种方法都保持了算子学习的性质,因此具备在任意分辨率下进行预测的能力。将我们的层添加到傅里叶神经算子中可显著提升其性能,在我们的实验中,相对 L2 误差降低了 34-72%,这些实验包括湍流二维 Navier-Stokes 方程和球面浅水方程。

英文摘要:

Neural operators learn mappings between function spaces, which is practical for learning solution operators of PDEs and other scientific modeling applications. Among them, the Fourier neural operator (FNO) is a popular architecture that performs global convolutions in the Fourier space. However, such global operations are often prone to over-smoothing and may fail to capture local details. In contrast, convolutional neural networks (CNN) can capture local features but are limited to training and inference at a single resolution. In this work, we present a principled approach to operator learning that can capture local features under two frameworks by learning differential operators and integral operators with locally supported kernels. Specifically, inspired by stencil methods, we prove that we obtain differential operators under an appropriate scaling of the kernel values of CNNs. To obtain local integral operators, we utilize suitable basis representations for the kernels based on discrete-continuous convolutions. Both these approaches preserve the properties of operator learning and, hence, the ability to predict at any resolution. Adding our layers to FNOs significantly improves their performance, reducing the relative L2-error by 34-72% in our experiments, which include a turbulent 2D Navier-Stokes and the spherical shallow water equations.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑