Overview

I am investigating neural-operator surrogates for high-frequency RF wave propagation in tokamak plasmas. A central challenge is learning highly oscillatory complex fields: small spatial errors can become large phase errors even when the overall field amplitude appears plausible.

Research direction

This ongoing work explores physics-aligned representations for improving the fidelity and robustness of surrogate models in regimes that are difficult for direct real-space operator learning. This public page focuses on the research problem and current status; technical details will accompany the publication.

The method is currently being prepared for publication.