Baptiste Mathevon, Farah Cherfaoui, Amaury Habrard, Marc Sebban
This paper proposes a taxonomy of PDE evaluation metrics and discusses their advantages and limitations. It aims to promote standardized and reliable practices in PDE discovery. The paper provides recommendations for ML experts and users of PDE discovery methods.
The evaluation of discovered PDEs is crucial for scientific advancement and real-world applications. Standardized practices can prevent overly interpreted conclusions and ensure the validity of new physical theories. Reliable PDE discovery methods can lead to breakthroughs in various fields, including physics and engineering.
Thank you to arXiv for use of its open access interoperability. arxiv.org. aibrief.ai is not affiliated with or endorsed by arXiv.