Research article

Physics-informed neural networks for reaction-diffusion systems: a comparative study with numerical integration of the FitzHugh–Nagumo equations

  • Published: 19 November 2026
  • This study employs physics-informed neural networks (PINNs) and the explicit Euler, second-order Runge–Kutta (RK2) and fourth-order Runge–Kutta (RK4) methods to solve the two-dimensional FitzHugh–Nagumo (FHN) reaction-diffusion system with periodic boundary conditions. The PINN combines the governing equations, the initial condition, and sparse samples of the RK4 solution in its loss function, and all methods are evaluated against a high-accuracy pseudo-spectral reference solution on test points that are not used for training. The results show that the PINN reproduces the spatiotemporal evolution of the system at grid and off-grid points. Owing to the physics loss, the PINN achieves lower errors than RK4 for the excitation variable and maintains its accuracy with very sparse data, although it requires much longer training and is less accurate for the recovery variable. It is therefore a promising complement to classical solvers, particularly when data are sparse.

    Citation: Chao Li, Xiyuan Xiao, Yan Du, Lingxi Xia. Physics-informed neural networks for reaction-diffusion systems: a comparative study with numerical integration of the FitzHugh–Nagumo equations[J]. Big Data and Information Analytics, 2026, 10: 391-408. doi: 10.3934/bdia.2026020

    Related Papers:

  • This study employs physics-informed neural networks (PINNs) and the explicit Euler, second-order Runge–Kutta (RK2) and fourth-order Runge–Kutta (RK4) methods to solve the two-dimensional FitzHugh–Nagumo (FHN) reaction-diffusion system with periodic boundary conditions. The PINN combines the governing equations, the initial condition, and sparse samples of the RK4 solution in its loss function, and all methods are evaluated against a high-accuracy pseudo-spectral reference solution on test points that are not used for training. The results show that the PINN reproduces the spatiotemporal evolution of the system at grid and off-grid points. Owing to the physics loss, the PINN achieves lower errors than RK4 for the excitation variable and maintains its accuracy with very sparse data, although it requires much longer training and is less accurate for the recovery variable. It is therefore a promising complement to classical solvers, particularly when data are sparse.



    加载中


    [1]

    Polyanin AD, Sorokin VG, Vyazmin AV, (2018) Reaction-diffusion models with delay: Some properties, equations, problems, and solutions. Theor Found Chem Eng 52: 334–348. https://doi.org/10.1134/S0040579518030132

    doi: 10.1134/S0040579518030132
    [2]

    FitzHugh R, (1961) Impulses and physiological states in theoretical models of nerve membrane. Biophys J 1: 445–466. https://doi.org/10.1016/S0006-3495(61)86902-6

    doi: 10.1016/S0006-3495(61)86902-6
    [3]

    Nagumo J, Arimoto S, Yoshizawa S, (1962) An active pulse transmission line simulating nerve axon. Proc IRE 50: 2061–2070. https://doi.org/10.1109/JRPROC.1962.288235

    doi: 10.1109/JRPROC.1962.288235
    [4]

    Ilati M, Dehghan M, (2017) Application of direct meshless local Petrov-Galerkin (DMLPG) method for some Turing-type models. Eng Comput 33: 107–124. https://doi.org/10.1007/s00366-016-0458-x

    doi: 10.1007/s00366-016-0458-x
    [5]

    Nishiyama H, Noi T, (2016) Conservative difference schemes for the numerical solution of the Gardner equation. Comput Appl Math 35: 75–95. https://doi.org/10.1007/s40314-014-0183-2

    doi: 10.1007/s40314-014-0183-2
    [6]

    Kaiser MW, Fries TP, (2025) Simultaneous solution of incompressible Navier-Stokes flows on multiple surfaces. Arch Appl Mech 95: 233. https://doi.org/10.1007/s00419-025-02935-z

    doi: 10.1007/s00419-025-02935-z
    [7]

    DiBenedetto E, Gianazza U, (2023) Partial Differential Equations, 3rd Ed., Cham: Birkhäuser. https://doi.org/10.1007/978-3-031-46618-2

    [8]

    Raissi M, Perdikaris P, Karniadakis GE, (2019) Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J Comput Phys 378: 686–707. https://doi.org/10.1016/j.jcp.2018.10.045

    doi: 10.1016/j.jcp.2018.10.045
    [9]

    Hu B, McDaniel D, (2023) Applying physics-informed neural networks to solve Navier-Stokes equations for laminar flow around a particle. Math Comput Appl 28: 102. https://doi.org/10.3390/mca28050102

    doi: 10.3390/mca28050102
    [10]

    Bai J, Rabczuk T, Gupta A, Alzubaidi L, Gu Y, (2023) A physics-informed neural network technique based on a modified loss function for computational 2D and 3D solid mechanics. Comput Mech 71: 543–562. https://doi.org/10.1007/s00466-022-02252-0

    doi: 10.1007/s00466-022-02252-0
    [11]

    Sarabian M, Babaee H, Laksari K, (2022) Physics-informed neural networks for brain hemodynamic predictions using medical imaging. IEEE Trans Med Imaging 41: 2285–2303. https://doi.org/10.1109/TMI.2022.3161653

    doi: 10.1109/TMI.2022.3161653
    [12]

    Karniadakis GE, Kevrekidis IG, Lu L, Perdikaris P, Wang S, Yang L, (2021) Physics-informed machine learning. Nat Rev Phys 3: 422–440. https://doi.org/10.1038/s42254-021-00314-5

    doi: 10.1038/s42254-021-00314-5
    [13]

    Krishnapriyan AS, Gholami A, Zhe S, Kirby RM, Mahoney MW, (2021) Characterizing possible failure modes in physics-informed neural networks. Adv Neural Inf Process Syst 34: 26548–26560..

    [14]

    Wang S, Teng Y, Perdikaris P, (2021) Understanding and mitigating gradient flow pathologies in physics-informed neural networks. SIAM J Sci Comput 43: A3055–A3081. https://doi.org/10.1137/20M1318043

    doi: 10.1137/20M1318043
    [15]

    Ren P, Rao C, Liu Y, Wang JX, Sun H, (2022) PhyCRNet: Physics-informed convolutional-recurrent network for solving spatiotemporal PDEs. Comput Methods Appl Mech Eng 389: 114399. https://doi.org/10.1016/j.cma.2021.114399

    doi: 10.1016/j.cma.2021.114399
    [16]

    Rao C, Ren P, Wang Q, Buyukozturk O, Sun H, Liu Y, (2023) Encoding physics to learn reaction-diffusion processes. Nat Mach Intell 5: 765–779. https://doi.org/10.1038/s42256-023-00685-7

    doi: 10.1038/s42256-023-00685-7
    [17]

    Xia L, Han B, (2025) Comparative study of PINNs and numerical methods in infectious disease reaction-diffusion models. Big Data Inf Anal 9: 265–284. https://doi.org/10.3934/bdia.2025012

    doi: 10.3934/bdia.2025012
    [18]

    Dong S, Ni N, (2021) A method for representing periodic functions and enforcing exactly periodic boundary conditions with deep neural networks. J Comput Phys 435: 110242. https://doi.org/10.1016/j.jcp.2021.110242

    doi: 10.1016/j.jcp.2021.110242
    [19]

    Wang S, Sankaran S, Perdikaris P, (2024) Respecting causality for training physics-informed neural networks. Comput Methods Appl Mech Eng 421: 116813. https://doi.org/10.1016/j.cma.2024.116813

    doi: 10.1016/j.cma.2024.116813
    [20]

    Wu C, Zhu M, Tan Q, Kartha Y, Lu L, (2023) A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks. Comput Methods Appl Mech Eng 403: 115671. https://doi.org/10.1016/j.cma.2022.115671

    doi: 10.1016/j.cma.2022.115671
    [21]

    Garvie MR, (2007) Finite-difference schemes for reaction-diffusion equations modeling predator-prey interactions in MATLAB. Bull Math Biol 69: 931–956. https://doi.org/10.1007/s11538-006-9062-3

    doi: 10.1007/s11538-006-9062-3
    [22]

    Arif MS, Abodayeh K, Nawaz Y, (2025) A hybrid finite difference approach for solving fuzzy stochastic SIR-$\beta$ model with diffusion and incidence rate. Eur J Pure Appl Math 18: 6292. https://doi.org/10.29020/nybg.ejpam.v18i3.6292

    doi: 10.29020/nybg.ejpam.v18i3.6292
    [23]

    Madzvamuse A, Chung AHW, (2014) Fully implicit time-stepping schemes and non-linear solvers for systems of reaction-diffusion equations. Appl Math Comput 244: 361–374. https://doi.org/10.1016/j.amc.2014.07.004

    doi: 10.1016/j.amc.2014.07.004
    [24]

    Boscarino S, Filbet F, Russo G, (2016) High order semi-implicit schemes for time dependent partial differential equations. J Sci Comput 68: 975–1001. https://doi.org/10.1007/s10915-016-0168-y

    doi: 10.1007/s10915-016-0168-y
    [25]

    Nie Q, Zhang YT, Zhao R, (2006) Efficient semi-implicit schemes for stiff systems. J Comput Phys 214: 521–537. https://doi.org/10.1016/j.jcp.2005.09.030

    doi: 10.1016/j.jcp.2005.09.030
    [26]

    Salah M, Matbuly MS, Civalek O, Ragb O, (2024) Calculation of four-dimensional unsteady gas flow via different quadrature schemes and Runge–Kutta 4th order method. Adv Appl Math Mech 16: 437–458. https://doi.org/10.4208/aamm.OA-2021-0373

    doi: 10.4208/aamm.OA-2021-0373
    [27]

    Kelleci A, Yıldırım A, (2011) Numerical solution of the system of nonlinear ordinary differential equations arising in kinetic modeling of lactic acid fermentation and epidemic model. Int J Numer Methods Biomed Eng 27: 585–594. https://doi.org/10.1002/cnm.1321

    doi: 10.1002/cnm.1321
    [28]

    Zhang Q, Wu C, Kahana A, Karniadakis GE, Kim Y, Li Y, et al. (2025) Artificial to spiking neural networks conversion with calibration in scientific machine learning. SIAM J Sci Comput 47: C559–C577. https://doi.org/10.1137/24M1643232

    doi: 10.1137/24M1643232
    [29]

    Jiang Q, Gou Z, (2025) Solutions to two- and three-dimensional incompressible flow fields leveraging a physics-informed deep learning framework and Kolmogorov-Arnold networks. Int J Numer Methods Fluids 97: 665–673. https://doi.org/10.1002/fld.5374

    doi: 10.1002/fld.5374
    [30]

    Cox SM, Matthews PC, (2002) Exponential time differencing for stiff systems. J Comput Phys 176: 430–455. https://doi.org/10.1006/jcph.2002.6995

    doi: 10.1006/jcph.2002.6995
    [31]

    Kassam AK, Trefethen LN, (2005) Fourth-order time-stepping for stiff PDEs. SIAM J Sci Comput 26: 1214–1233. https://doi.org/10.1137/S1064827502410633

    doi: 10.1137/S1064827502410633
    [32]

    LeVeque RJ, (2007) Finite Difference Methods for Ordinary and Partial Differential Equations: Steady-State and Time-Dependent Problems, Philadelphia: SIAM. https://doi.org/10.1137/1.9780898717839

    [33]

    Miyato T, Kataoka T, Koyama M, Yoshida Y, (2018) Spectral normalization for generative adversarial networks. In: International Conference on Learning Representations (ICLR).

    [34]

    Daw A, Bu J, Wang S, Perdikaris P, Karpatne A, (2023) Mitigating propagation failures in physics-informed neural networks using retain-resample-release (R3) sampling. In: Proceedings of the 40th International Conference on Machine Learning, 202: 7264–7302.

  • Reader Comments
  • © 2026 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)

Metrics

Article views(6) PDF downloads(1) Cited by(0)

Article outline

Figures and Tables

Figures(8)  /  Tables(6)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog