Heteroscedastic Neural Surrogate Modeling for Robust and Rapid Bayesian Inference in Fusion Plasma Diagnostics
This work proposes a neural-network-based probabilistic surrogate framework for rapid and robust MCMC inference that achieves>1500x acceleration over exact physical forward models, while simultaneously reducing inference error (RMSE) by>20% compared to standard homoscedastic neural baselines, offering a highly promisin...