Geodesic
Training a neural network for a hyperbolic equation by using a quasiclassical functional
Itogi nauki i tehniki. Sovremennaâ matematika i eë priloženiâ. Tematičeskie obzory, Proceedings of the Voronezh international spring mathematical school "Modern methods of the theory of boundary-value problems. Pontryagin readings—XXXV", Voronezh, April 26-30, 2024, Part 3, Tome 237 (2024), pp. 76-86.

Voir la notice de l'article provenant de la source Math-Net.Ru

We study the problem of constructing a loss functional based on the quasiclassical variational principle for training a neural network, which approximates solutions of a hyperbolic equation. Using the method of symmetrizing operator proposed by V. M. Shalov, for the second-order hyperbolic equation, we construct a variational functional of the boundary-value problem, which involves integrals over the domain of the boundary-value problem and a segment of the boundary, depending on first-order derivatives of the unknown function. We demonstrate that the neural network approximating the solution of the boundary-value problem considered can be trained by using the constructed variational functional.
Keywords: variational principle, hyperbolic equation, neural network, loss functional 
@article{INTO_2024_237_a5,
     author = {S. G. Shorokhov},
     title = {Training a neural network for a hyperbolic equation by using a quasiclassical functional},
     journal = {Itogi nauki i tehniki. Sovremenna\^a matematika i e\"e prilo\v{z}eni\^a. Temati\v{c}eskie obzory},
     pages = {76--86},
     publisher = {mathdoc},
     volume = {237},
     year = {2024},
     language = {ru},
     url = {http://geodesic.mathdoc.fr/item/INTO_2024_237_a5/}
}
TY  - JOUR
AU  - S. G. Shorokhov
TI  - Training a neural network for a hyperbolic equation by using a quasiclassical functional
JO  - Itogi nauki i tehniki. Sovremennaâ matematika i eë priloženiâ. Tematičeskie obzory
PY  - 2024
SP  - 76
EP  - 86
VL  - 237
PB  - mathdoc
UR  - http://geodesic.mathdoc.fr/item/INTO_2024_237_a5/
LA  - ru
ID  - INTO_2024_237_a5
ER  - 
%0 Journal Article
%A S. G. Shorokhov
%T Training a neural network for a hyperbolic equation by using a quasiclassical functional
%J Itogi nauki i tehniki. Sovremennaâ matematika i eë priloženiâ. Tematičeskie obzory
%D 2024
%P 76-86
%V 237
%I mathdoc
%U http://geodesic.mathdoc.fr/item/INTO_2024_237_a5/
%G ru
%F INTO_2024_237_a5
S. G. Shorokhov. Training a neural network for a hyperbolic equation by using a quasiclassical functional. Itogi nauki i tehniki. Sovremennaâ matematika i eë priloženiâ. Tematičeskie obzory, Proceedings of the Voronezh international spring mathematical school "Modern methods of the theory of boundary-value problems. Pontryagin readings—XXXV", Voronezh, April 26-30, 2024, Part 3, Tome 237 (2024), pp. 76-86. http://geodesic.mathdoc.fr/item/INTO_2024_237_a5/

[1] Sobol I. M., Chislennye metody Monte-Karlo, Nauka, M., 1973

[2] Filippov V. M., “Variatsionnyi metod resheniya kraevykh zadach dlya volnovogo uravneniya”, Differ. uravn., 20:11 (1984), 1961–1968

[3] Filippov V. M., Savchin V. M., Shorokhov S. G., “Variatsionnye printsipy dlya nepotentsialnykh operatorov”, Itogi nauki tekhn. Sovr. probl. mat. Nov. dostizh., 40 (1992), 3–176

[4] Shalov V. M., “Nekotoroe obobschenie prostranstva K. Fridrikhsa”, Dokl. AN SSSR., 151:2 (1963), 292–294

[5] Shalov V. M., “Reshenie nesamosopryazhennykh uravnenii variatsionnym metodom”, Dokl. AN SSSR., 151:3 (1963), 511–512

[6] Shalov V. M., “Printsip minimuma kvadratichnogo funktsionala dlya giperbolicheskogo uravneniya”, Differ. uravn., 1:10 (1965), 1338–1365

[7] E W., Yu B., “The deep Ritz method: A deep learning-based numerical algorithm for solving variational problems”, Commun. Math. Stat., 6:1 (2018), 1–12

[8] Geneva N., Zabaras N., “Modeling the dynamics of PDE systems with physics-constrained deep auto-regressive networks”, J. Comput. Phys., 403 (2020), 109056

[9] Raissi M., Perdikaris P., Karniadakis G., “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations”, J. Comput. Phys., 378 (2019), 686–707

[10] Sirignano J., Spiliopoulos K., “DGM: A deep learning algorithm for solving partial differential equations”, J. Comput. Phys., 375 (2018), 1339–1364

[11] Young W. H., “On multiple integration by parts and the second theorem of the mean”, Proc. London Math. Soc., 2:16 (1917), 273–293

[12] Zhu Y., Zabaras N., Koutsourelakis P.-S., Perdikaris P., “Physics-constrained deep learning for high-dimensional surrogate modeling and uncertainty quantification without labeled data”, Journal of Computational Physics, 394 (2019), 56–81