清晨好,您是今天最早来到科研通的研友!由于当前在线用户较少,发布求助请尽量完整的填写文献信息,科研通机器人24小时在线,伴您科研之路漫漫前行!

A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks

采样(信号处理) 自适应采样 残余物 重采样 算法 拉丁超立方体抽样 人工神经网络 计算机科学 序列(生物学) 切片取样 数学 数学优化 重要性抽样 人工智能 统计 蒙特卡罗方法 滤波器(信号处理) 生物 遗传学 计算机视觉
作者
Chenxi Wu,Min Zhu,Qinyang Tan,Yadhu Kartha,Lu Lu
出处
期刊:Computer Methods in Applied Mechanics and Engineering [Elsevier]
卷期号:403: 115671-115671 被引量:194
标识
DOI:10.1016/j.cma.2022.115671
摘要

Physics-informed neural networks (PINNs) have shown to be an effective tool for solving forward and inverse problems of partial differential equations (PDEs). PINNs embed the PDEs into the loss of the neural network, and this PDE loss is evaluated at a set of scattered residual points. The distribution of these points are highly important to the performance of PINNs. However, in the existing studies on PINNs, only a few simple residual point sampling methods have mainly been used. Here, we present a comprehensive study of two categories of sampling: non-adaptive uniform sampling and adaptive nonuniform sampling. We consider six uniform sampling, including (1) equispaced uniform grid, (2) uniformly random sampling, (3) Latin hypercube sampling, (4) Halton sequence, (5) Hammersley sequence, and (6) Sobol sequence. We also consider a resampling strategy for uniform sampling. To improve the sampling efficiency and the accuracy of PINNs, we propose two new residual-based adaptive sampling methods: residual-based adaptive distribution (RAD) and residual-based adaptive refinement with distribution (RAR-D), which dynamically improve the distribution of residual points based on the PDE residuals during training. Hence, we have considered a total of 10 different sampling methods, including six non-adaptive uniform sampling, uniform sampling with resampling, two proposed adaptive sampling, and an existing adaptive sampling. We extensively tested the performance of these sampling methods for four forward problems and two inverse problems in many setups. Our numerical results presented in this study are summarized from more than 6000 simulations of PINNs. We show that the proposed adaptive sampling methods of RAD and RAR-D significantly improve the accuracy of PINNs with fewer residual points. The results obtained in this study can also be used as a practical guideline in choosing sampling methods.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
更新
大幅提高文件上传限制,最高150M (2024-4-1)

科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
Herbs完成签到 ,获得积分10
1秒前
hhhhhha完成签到,获得积分10
3秒前
15秒前
煜琪完成签到 ,获得积分10
15秒前
wyt发布了新的文献求助10
20秒前
光亮的自行车完成签到 ,获得积分10
20秒前
巫马白亦完成签到,获得积分10
24秒前
Hiaoliem完成签到 ,获得积分10
24秒前
zhdjj完成签到 ,获得积分10
1分钟前
xfcy完成签到,获得积分0
1分钟前
紫陌完成签到,获得积分0
1分钟前
修水县1个科研人完成签到 ,获得积分10
1分钟前
zijingsy完成签到 ,获得积分10
1分钟前
xixi很困完成签到 ,获得积分10
1分钟前
marinemiao完成签到,获得积分10
1分钟前
萧水白应助marinemiao采纳,获得10
1分钟前
mrwang完成签到 ,获得积分10
1分钟前
1分钟前
www发布了新的文献求助10
2分钟前
2分钟前
研友_VZG7GZ应助利酱采纳,获得10
2分钟前
dajiejie完成签到 ,获得积分10
2分钟前
勤劳的颤完成签到 ,获得积分10
2分钟前
2分钟前
利酱发布了新的文献求助10
2分钟前
franca2005完成签到 ,获得积分10
2分钟前
www完成签到,获得积分10
3分钟前
雪妮完成签到 ,获得积分10
3分钟前
migi完成签到,获得积分10
3分钟前
张大星完成签到 ,获得积分10
3分钟前
liuzhifenshen完成签到,获得积分10
4分钟前
elisa828完成签到,获得积分10
4分钟前
Jack80发布了新的文献求助50
4分钟前
huiluowork完成签到 ,获得积分10
4分钟前
5分钟前
5分钟前
fff发布了新的文献求助10
5分钟前
loga80完成签到,获得积分10
5分钟前
二牛完成签到,获得积分10
5分钟前
星希完成签到 ,获得积分10
5分钟前
高分求助中
Sustainability in Tides Chemistry 1500
Handbook of the Mammals of the World – Volume 3: Primates 805
拟南芥模式识别受体参与调控抗病蛋白介导的ETI免疫反应的机制研究 550
Gerard de Lairesse : an artist between stage and studio 500
Digging and Dealing in Eighteenth-Century Rome 500
Queer Politics in Times of New Authoritarianisms: Popular Culture in South Asia 500
Manual of Sewer Condition Classification 500
热门求助领域 (近24小时)
化学 医学 生物 材料科学 工程类 有机化学 生物化学 物理 内科学 纳米技术 计算机科学 化学工程 复合材料 基因 遗传学 催化作用 物理化学 免疫学 量子力学 细胞生物学
热门帖子
关注 科研通微信公众号,转发送积分 3068193
求助须知:如何正确求助?哪些是违规求助? 2722162
关于积分的说明 7476072
捐赠科研通 2369138
什么是DOI,文献DOI怎么找? 1256228
科研通“疑难数据库(出版商)”最低求助积分说明 609518
版权声明 596835