|
|
TCAS-PINN: Physics-informed neural networks with a novel temporal causality-based adaptive sampling method |
Jia Guo, Haifeng Wang, Shilin Gu, and Chenping Hou† |
College of Science, National University of Defense Technology, Changsha 410073, China |
|
|
Abstract Physics-informed neural networks (PINNs) have become an attractive machine learning framework for obtaining solutions to partial differential equations (PDEs). PINNs embed initial, boundary, and PDE constraints into the loss function. The performance of PINNs is generally affected by both training and sampling. Specifically, training methods focus on how to overcome the training difficulties caused by the special PDE residual loss of PINNs, and sampling methods are concerned with the location and distribution of the sampling points upon which evaluations of PDE residual loss are accomplished. However, a common problem among these original PINNs is that they omit special temporal information utilization during the training or sampling stages when dealing with an important PDE category, namely, time-dependent PDEs, where temporal information plays a key role in the algorithms used. There is one method, called Causal PINN, that considers temporal causality at the training level but not special temporal utilization at the sampling level. Incorporating temporal knowledge into sampling remains to be studied. To fill this gap, we propose a novel temporal causality-based adaptive sampling method that dynamically determines the sampling ratio according to both PDE residual and temporal causality. By designing a sampling ratio determined by both residual loss and temporal causality to control the number and location of sampled points in each temporal sub-domain, we provide a practical solution by incorporating temporal information into sampling. Numerical experiments of several nonlinear time-dependent PDEs, including the Cahn-Hilliard, Korteweg-de Vries, Allen-Cahn and wave equations, show that our proposed sampling method can improve the performance. We demonstrate that using such a relatively simple sampling method can improve prediction performance by up to two orders of magnitude compared with the results from other methods, especially when points are limited.
|
Received: 18 October 2023
Revised: 08 January 2024
Accepted manuscript online:
|
PACS:
|
07.05.Mh
|
(Neural networks, fuzzy logic, artificial intelligence)
|
|
02.60.Cb
|
(Numerical simulation; solution of equations)
|
|
02.30.Jr
|
(Partial differential equations)
|
|
84.35.+i
|
(Neural networks)
|
|
Fund: Project supported by the Key National Natural Science Foundation of China (Grant No. 62136005) and the National Natural Science Foundation of China (Grant Nos. 61922087, 61906201, and 62006238). |
Corresponding Authors:
Chenping Hou,E-mail:hcpnudt@hotmail.com
E-mail: hcpnudt@hotmail.com
|
Cite this article:
Jia Guo, Haifeng Wang, Shilin Gu, and Chenping Hou TCAS-PINN: Physics-informed neural networks with a novel temporal causality-based adaptive sampling method 2024 Chin. Phys. B 33 050701
|
[1] Karniadakis G E, Kevrekidis I G, Lu L, Perdikaris P, Wang S F and Liu Y 2021 Nat. Rev. Phys. 3 422 [2] Cuomo S, Di Cola V S, Giampaolo F, Rozza Z, Raissi M and Piccialli F 2022 J. Sci. Comput. 92 88 [3] Hao Z K, Liu S M, Zhang Y C, Ying C Y, Feng Y, Su H and Zhu J 2023 arXiv:2211.08064 [cs.LG] [4] Chen Y T and Zhang D X 2022 arXiv:2202.10337 [cs.AI] [5] E W N 2020 Commun. Comput. Phys. 28 1639 [6] Carleo G and Matthias T 2017 Science 355 602 [7] Arridge S, Maass P, Öktem O and Schönlieb C B 2019 Acta Numerica 28 1 [8] Karpatne A, Atluri G, Faghmous J H, Steinbach M, Banerjee A, Ganguly A, Shekhar S, Samatova N and Kumar V 2017 IEEE Trans. Knowledge Data Eng. 29 2318 [9] Huerta E A, Khan A, Huang X B, Tian M Y, Levental M, Chard R, Wei W, Heflin M, Katz D S, Kindratenko V, Mu D, Blaiszik B and Foster I 2021 Nat. Astron. 5 1062 [10] Meng X F, Hao X L, Ma C H, Yang C, Maoliniyazi A S M, Wu C and Wei J Y 2023 Chinese Journal of Computers 46 877 (in Chinese) [11] Raissi M, Perdikaris P and Karniadakis G E 2019 J. Comput. Phys. 378 686 [12] Jagtap A D, Kawaguchi K and Karniadakis G E 2020 J. Comput. Phys. 404 109136 [13] Shukla K, Jagtap A D and Karniadakis G E 2021 J. Comput. Phys. 447 110683 [14] Liu Y, Meng X H and Karniadakis G E 2021 J. Comput. Phys. 425 109913 [15] Pang G F, Lu L and Karniadakis G E 2019 SIAM J. Sci. Comput. 41 A2603 [16] Jagtap A D and Karniadakis G E 2020 Commun. Comput. Phys. 28 2002 [17] Krishnapriyan A, Gholami A, Zhe S D, Kirby R and Mahoney M W 2021 Proceedings of the 35th Advances in Neural Information Processing Systems, December 6–14, 2021, virtual-only, pp. 26548–26560 [18] Gao Z W, Yan L and Zhou T 2023 SIAM J. Sci. Comput. 45 A1971 [19] Peng W, Zhou W E, Zhang X Y, Yao W J and Liu Z L 2022 arXiv:2205.01051 [cs.LG] [20] Wang S F, Sankaran S and Perdikaris P 2022 arXiv:2203.07404 [cs.LG] [21] Gao Z H, Tang T, Yan L and Zhou T 2023 2023 Commun. Appl. Math. Comput. [22] Nabian M A, Gladstone R J and Meidani H 2021 Computer-Aided Civil and Infrastructure Engineering 36 962 [23] Mattey R and Ghosh S 2022 Comput. Methods Appl. Mech. Eng. 390 114474 [24] Wight C L and Zhao J 2021 Commun. Comput. Phys. 29 930 [25] Liu J N, Hou Q Z, Wei J G and Sun Z W 2023 Chin. Phys. B 32 070702 [26] Kuramoto Y and Tsuzuki T 1976 Progress of Theoretical Physics 55 356 [27] Michelson D M and Sivashinsky G I 1977 Acta Astronautica 4 1207 [28] Hyman J M and Nicolaenko B 1986 Physica D 18 113 [29] Jameson A, Schmidt W and Turkel E 2012 Proceedings of the 14th AIAA Fluid and Plasma Dynamics Conference, June 23–25, 1981, Palo Alto, CA, USA, p. 15 [30] Cox S M and Matthews P C 2002 J. Comput. Phys. 176 430 [31] Chen T P and Chen H 1995 IEEE Trans. Neural Networks 6 911 [32] Baydin A G, Pearlmutter B A, Radul A A and Siskind J M 2017 J. Mach. Learn. Res. 18 5595 [33] Lu L, Meng X H, Mao Z P and Karniadakis G E 2021 SIAM Review 63 208 [34] Stein M L 1987 Technometrics 29 143 [35] Blazek J 2015 Computational Fluid Dynamics: Principles and Applications (Third Edition) (Oxford) [36] Driscoll T A, Hale N and Trefethen L N 2014 Chebfun Guide (Oxford) [37] Dong S C and Ni N X 2021 J. Comput. Phys. 435 110242 [38] McClenny L D and Braga-Neto U M 2023 J. Comput. Phys. 474 111722 [39] Sun J W, Qian X, Zhang H and Song S H 2021 Chin. Phys. B 30 070201 [40] Tang K J, Wan X L and Yang C 2023 J. Comput. Phys. 476 111868 [41] Wang X L, Zhou T and Zhou Y C 2023 arXiv:2310.17185 [cs.LG] |
No Suggested Reading articles found! |
|
|
Viewed |
|
|
|
Full text
|
|
|
|
|
Abstract
|
|
|
|
|
Cited |
|
|
|
|
Altmetric
|
blogs
Facebook pages
Wikipedia page
Google+ users
|
Online attention
Altmetric calculates a score based on the online attention an article receives. Each coloured thread in the circle represents a different type of online attention. The number in the centre is the Altmetric score. Social media and mainstream news media are the main sources that calculate the score. Reference managers such as Mendeley are also tracked but do not contribute to the score. Older articles often score higher because they have had more time to get noticed. To account for this, Altmetric has included the context data for other articles of a similar age.
View more on Altmetrics
|
|
|