BubbleOKAN скачать в хорошем качестве

BubbleOKAN

BubbleOKAN is a physics‑informed neural operator that maps pressure profiles to high‑frequency bubble radius dynamics using a two‑step DeepOKAN (Kolmogorov–Arnold networks) and spline‑based trunk functions to mitigate spectral bias. It enhances interpretability and accurately captures both low‑ and high‑frequency behaviors governed by Rayleigh‑Plesset and Keller‑Miksis equations. This makes it a scalable, interpretable alternative to traditional numerical solvers and outperforms state‑of‑the‑art neural operators in benchmark scenarios.