Silke Glas (Computer Science, Cornell University)
In this talk, we consider the optimization of stellarators, which are a class of devices to generate fusion energy. We present the status quo of stellarator optimization and the arising challenges within this formulation. One particular challenge is, that slight errors in the building process can lead to a significant loss of generated fusion energy. In order to achieve a more stable configuration, we introduce a surrogate-based method modelling the uncertainty as input parameter.