Lyapunov weights to convey the meaning of time in physics-informed neural networks
Abstract: Time is not a dimension as the others. In Physics-Informed Neural Networks (PINN) several proposals attempted to adapt the time sampling or time weighting to take into account the specifics of this special dimension. But these proposals are not principled and need guidance to be used. We explain here theoretically why the Lyapunov exponents give actionable insights and propose a weighting scheme to automatically adapt to chaotic, periodic or stable dynamics. We characterize theoretically the best weighting scheme under computational constraints as a cumulative exponential integral of the local Lyapunov exponent estimators and show that it performs well in practice under the regimes mentioned above.
- doi:https://doi.org/10.1016/j.jcp.2018.10.045. URL https://www.sciencedirect.com/science/article/pii/S0021999118307125
- arXiv:https://doi.org/10.1137/20M1318043, doi:10.1137/20M1318043. URL https://doi.org/10.1137/20M1318043
- arXiv:https://doi.org/10.1080/14697688.2023.2294799, doi:10.1080/14697688.2023.2294799. URL https://doi.org/10.1080/14697688.2023.2294799
- arXiv:2307.12306.
- doi:10.3390/math12060803. URL https://www.mdpi.com/2227-7390/12/6/803
- doi:https://doi.org/10.1016/j.jcp.2022.111722. URL https://www.sciencedirect.com/science/article/pii/S0021999122007859
- doi:https://doi.org/10.1016/j.cma.2024.116813. URL https://www.sciencedirect.com/science/article/pii/S0045782524000690
- doi:https://doi.org/10.1016/j.jcp.2023.112464. URL https://www.sciencedirect.com/science/article/pii/S0021999123005594
- G. Turinici, Optimal time sampling in physics-informed neural networks (2024). arXiv:2404.18780.
- doi:https://doi.org/10.1016/0045-7930(86)90036-8. URL https://www.sciencedirect.com/science/article/pii/0045793086900368
- arXiv:https://doi.org/10.1137/0705041, doi:10.1137/0705041. URL https://doi.org/10.1137/0705041
- doi:https://doi.org/10.1016/B978-0-12-358502-8.50019-0. URL https://www.sciencedirect.com/science/article/pii/B9780123585028500190
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.