| Original language | English |
|---|---|
| Publication status | Published - 1 Jan 2019 |
| Event | 46th European Physical Society Conference on Plasma Physics (EPS 2019) - Milan, Italy Duration: 8 Jul 2019 → 12 Jul 2019 Conference number: 46 |
Conference
| Conference | 46th European Physical Society Conference on Plasma Physics (EPS 2019) |
|---|---|
| Abbreviated title | EPS 2019 |
| Country/Territory | Italy |
| City | Milan |
| Period | 8/07/19 → 12/07/19 |
Funding
A 15D training set of ∼ 105 individual points for the training of a NN surrogate model of QuaLiKiz was successfully constructed. By sampling the profiles fitted using the GPR algorithm on experimental data, the data set remains within an experimentally relevant parameter space. Preliminary NNs generated with this reduced data set are in good agreement with the base model and with previous NNs in regions of parameter space where the current data set density is sufficiently high. In order to increase the data set density further to improve the NN accuracy, these gradient quantities of the extracted data points will be expanded according to their uncertainties provided by the fitting routine. Acknowledgements This work has been carried out within the framework of the EUROfusion Consortium and has received funding from the Euratom research and training programme 2014-2018 and 2019-2020 under grant agreement No 633053 and from the RCUK [grant number EP/P012450/1].The views and opinions expressed herein do not necessarily reflect those of the European Commission.