VICTOR (Variational Inference for Confined Tokamak Output Reconstruction) — A Physics-Informed Neural Network that reconstructs plasma emissivity on the WEST Tokamak in ~0.5ms, trained entirely without ground-truth labels. Enabling real-time fusion reactor diagnostics and disruption prevention.
flax tokamak nuclear-fusion open-innovation jax google-cloud-run plasma-diagnostics streamlit physics-informed-neural-network vertex-ai sdg7 soft-xray-tomography unbiased-ai gsc2026 deep-learning-scientific-computing
-
Updated
Apr 23, 2026 - Jupyter Notebook