Visual data science to master complex simulation ensembles

Spatiotemporal ensembles often result from physical simulations. These ensembles contain many information-rich members, each corresponding to different simulation input parameters. The extensive data size makes manual analysis infeasible, necessitating automated approaches to assist the analysis. In the preceding project (PN 6-8 (I)), methods and tools were developed to assist the visual analysis of spatiotemporal ensemble data, such as determining member similarities or sensitivities of simulation parameters. In this project, we aim to generalize these methods and improve their resource efficiency to reduce computational costs when applied to new problems and data. For this, we explore techniques such as transfer learning and investigate how incorporating additional information, like attention data during the interactive analysis, can improve our methods. Furthermore, we research how these methods can be applied in immersive environments such as virtual reality, which offers additional freedom for interacting with the environment and visualizing data.