Skip to main content
Skip header

Artificial intellignece in physics

Type of study Follow-up Master
Language of instruction English
Code 653-3177/02
Abbreviation UIF
Course title Artificial intellignece in physics
Credits 3
Coordinating department Department of Materials Engineering and Recycling
Course coordinator Ing. Dalibor Javůrek, Ph.D.

Subject syllabus

1. Usage of neural networks in science, past and present.
2. Structure of the solution of a computational problem with machine learning: Identification of the problem; data acquisition, sorting, and processing of data and their connection with a model; correct choice of an architecture; loss function; optimization - training of the model; acceleration of training, regularization.
3. Utilization of deep learning in computational methods, architectures of neural networks: Hamiltonian neural networks, Fourier neural operator, physically informed neural networks, convolutional neural networks, graph neural networks.
4. Autoencoders and dimensionality reduction of the acquired data.
5. Identification of physical laws from experimental data.
6. Numerical computation with deep learning: Quadrature problem.
7. Physically informed neural networks (PINNs).
8. PINNs and their versions – Loss re-weighting and data resampling, optimization targets.
9. PINNs and their versions – Regularization techniques, new neural architectures, new paradigms in the training of PINNs, and future outlook.
10. PINNs and their versions – Advanced methods in the implementation of physical constraints: Loss function, optimization algorithm, architecture of the neural network.
11. Optimization algorithms in physical tasks.
12. Additional information for the course, discussion.

Literature

[1] Steve Brunton. Physics Informed Machine Learning. Accessed: 2024-08-01.
2024. url: https://www.youtube.com/playlist?list=PLMrJAkhIeNNQ0BaKuBKY43k4xMo6NSbBa.
[2] ETH Zürich CAMLab. ETH Zürich — Deep Learning in Scientific Compu-
ting 2023. Accessed: 2024-08-01. 2024. url: https://www.youtube.com/
watch?v=y6wHpRzhhkA&list=PLJkYEExhe7rYY5HjpIJbgo-tDZ3bIAqAm.
[3] Zhongkai Hao et al. Physics-Informed Machine Learning: A Survey on Pro-
blems, Methods and Applications. 2023. arXiv: 2211.08064 [cs.LG]. url:
https://arxiv.org/abs/2211.08064.
[4] Stefan Kollmannsberger et al. Deep Learning in Computational Mechanics:
An Introductory Course. Springer Cham, 2021. isbn: 978-3-030-76586-6 .
doi: 10.1007/978-3-030-76587-3.
[5] Karet Pentil. PINN (Physics-Informed Neural Networks). Accessed: 2024-
08-01. 2021. url: https://www.youtube.com/playlist?list=PLXmYoJbJ848pkMm9NGZZKXUQJ8XWIXZX8
[6] Nils Thuerey et al. Physics-based Deep Learning. 2022. arXiv: 2109.05237 [cs.LG].
url: https://arxiv.org/abs/2109.05237.
[7] Genki Yagawa a Atsuya Oishi. Computational Mechanics with Deep Lear-
ning: An Introduction. Lecture Notes on Numerical Methods in Enginee-
ring and Sciences. Springer Cham, 2022. isbn: 978-3-031-11846-3 . doi:
10.1007/978-3-031-11847-0.
[8] M. Raissi, P. Perdikaris a G.E. Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics, 2019. doi: 10.1016/j.jcp.2018.10.045.
[9] Edward Small. An Analysis of Physics-Informed Neural Networks. arXiv:2303.02890 [cs.LG], 2023. url: https://arxiv.org/abs/2303.02890

Advised literature

[1] Ricardo Vinuesa. Physics-informed neural networks for fluid mechanics.
Accessed: 2024-08-01. 2024. url: https://www.youtube.com/watch?v=EHrgSPHZG3Y.
[2] Timon Rabczuk and Klaus-Jürgen Bathe. Machine Learning in Modeling and Simulation: Methods and Applications. Springer Cham, 2023. isbn: 978-3-031-36643-6 . doi: 10.1007/978-3-031-36644-4.
[3] Steven L. Brunton a J. Nathan Kutz. Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control. Cambridge University Press, 2023. isbn: 978-1009098489.
[4] Dimitrios Angelis et al. Artificial Intelligence in Physical Sciences: Symbolic Regression Trends and Perspectives. Springer Cham, 2023. isbn: 978-3-030-99226-2 .