Autoregressive transformers for disruption prediction in nuclear fusion plasmas
Spangher, L., Arnold, W., Spangher, A., Maris, A.D., Rea, C.
Most disruption predictors classify individual time slices independently. This work instead treats a discharge as a sequence and uses an autoregressive transformer to forecast instability onset, which lets the model use the trajectory a plasma is on rather than only where it currently sits.
arXiv:2401.00051.
CitationSpangher, L., W. Arnold, A. Spangher, A. Maris, and C. Rea. "Autoregressive transformers for disruption prediction in nuclear fusion plasmas." arXiv:2401.00051 (2023).