Neural network-accelerated Bayesian inference of impurity transport in fusion plasmas

Understanding and control of impurity transport in fusion plasmas is crucial on the way to abundant, clean and safe fusion energy. Using powerful Bayesian inference, accelerated by neural network surrogate models, we will estimate impurity concentrations and parameters governing transport, including accompanying uncertainties. With these tools, impurity transport will be studied in several devices, preparing for impurity control in ITER.