Operator inference with roll outs for learning reduced models from scarce and low-quality data

Operator inference with roll outs for learning reduced models from scarce and low-quality data
复制标题

推出算子推理,从稀缺和低质量的数据中学习简化模型

DOI:
10.1016/j.camwa.2023.06.012
复制
发表时间:
2023
影响因子:
2.9
通讯作者:
Peherstorfer, Benjamin
Peherstorfer, Benjamin
中科院分区:
数学2区
文献类型:
--
作者:
Uy, Wayne Isaac;Hartmann, Dirk;Peherstorfer, Benjamin

文献摘要

参考文献

被引文献

相似文献

数据驱动建模已经成为计算科学和工程的一个重要组成部分。然而,科学和工程中可用的数据通常是稀缺的,经常被噪声污染,并受到测量误差和其他扰动的影响,这使得学习系统动力学具有挑战性。在这项工作中,我们建议将通过算子推理的数据驱动建模与通过神经常微分方程的推出的动态训练相结合。带rollout的算子推理继承了传统算子推理的可解释性、可扩展性和结构保留,同时利用通过多个时间步的rollout进行动态训练,以提高从低质量和噪声数据中学习的稳定性和鲁棒性。用描述浅水波浪和地表准地转动力学的数据进行的数值实验表明,即使数据在时间上是稀疏采样的,并且受到高达10%的噪声污染,带滚动的算子推理也可以从训练轨迹中提供预测模型。
Data-driven modeling has become a key building block in computational science and engineering. However, data that are available in science and engineering are typically scarce, often polluted with noise and affected by measurement errors and other perturbations, which makes learning the dynamics of systems challenging. In this work, we propose to combine data-driven modeling via operator inference with the dynamic training via roll outs of neural ordinary differential equations. Operator inference with roll outs inherits interpretability, scalability, and structure preservation of traditional operator inference while leveraging the dynamic training via roll outs over multiple time steps to increase stability and robustness for learning from low-quality and noisy data. Numerical experiments with data describing shallow water waves and surface quasi-geostrophic dynamics demonstrate that operator inference with roll outs provides predictive models from training trajectories even if data are sampled sparsely in time and polluted with noise of up to 10%.
通过数据驱动算子推理的参数偏微分方程的非侵入式降阶模型
DOI: --
发表时间: 2021
影响因子: 3.1
作者:
Shane A. McQuarrie;Parisa Khodabakhshi;K. Willcox
通讯作者: K. Willcox
二次系统的稳定性分析
DOI: 10.1016/b978-0-08-037022-4.50039-9
发表时间: 1989
期刊: 2021 European Control Conference (ECC)
影响因子: --
作者:
R. Genesio;A. Tesi
通讯作者: A. Tesi
DOI: 10.1137/15m1041432
发表时间: 2016-01-01
影响因子: 3.1
作者:
Antoulas, A. C.;Gosea, I. V.;Ionita, A. C.
通讯作者: Ionita, A. C.
DOI: 10.1016/j.cma.2020.113433
发表时间: 2020-12-01
影响因子: 7.2
作者:
Benner, Peter;Goyal, Pawan;Willcox, Karen
通讯作者: Willcox, Karen
DOI: 10.1109/nnsp.1994.366006
发表时间: 1994-09
期刊: Proceedings of IEEE Workshop on Neural Networks for Signal Processing
影响因子: --
作者:
R. Rico-Martínez;J. S. Anderson;I. Kevrekidis
通讯作者: R. Rico-Martínez;J. S. Anderson;I. Kevrekidis