Data from direct computation or numerical simulation, often high-dimensional, can be subject to various uncertainties including parameter or domain uncertainty, model uncertainty, numerical errors or some specific stochasticity of the model. The computed data is a representative sample of a certain mathematical distribution. Directly visualizing this distribution is a difficult task, because in most cases it is not known or computationally expensive to obtain. Thus, further uncertainties are added to the outcome by the use of e.g. filtering steps or abstraction and rendering techniques to help in the visualization.
Further, this might also be affected by the choice of a risk measure, an a-priori task to the simulation and visualization, which is not always feasible. The choice can change the whole sampling strategy, e.g. a rare-event analysis requires the application of some importance sampling techniques.
For further information please contact Cedric Beschle

Cedric Beschle
Dr.Former Research Assistant