.png){kind=icon}
返回
Learning to differentiate
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jcp.2020.109873.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Artificial neural networks together with associated computational libraries provide a powerful framework for constructing both classification and regression algorithms. In this paper we use neural networks to design linear and non-linear discrete differential operators. We show that neural network based operators can be used to construct stable discretizations of initial boundary-value problems by ensuring that the operators satisfy a discrete analogue of integration-by-parts known as summation-by-parts. Our neural network approach with linear activation functions is compared and contrasted with a more traditional linear algebra approach. An application to overlapping grids is explored. The strategy developed in this work opens the door for constructing stable differential operators on general meshes. (C) 2020 Elsevier Inc. All rights reserved.
Keyword:
Neural networks
Discrete differential operators
Stability
Summation-by-parts
Overlapping grids
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
IF:
3.8
论文数:
1.6W
被引数:
8.4W
机构
引用论文
Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations物理信息神经网络: 一种用于解决涉及非线性偏微分方程的正反问题的深度学习框架
OAAI
OAAI分享收藏A unified deep artificial neural network approach to partial differential equations in complex geometries复杂几何中偏微分方程的统一深度人工神经网络方法.png)
NEUROCOMPUTINGOAAI
NEUROCOMPUTINGIF6.5
OAAI分享收藏
查看更多内容