SC²S Colloquium - October 28, 2015
|Date:||October 28, 2015|
|Time:||3:00 pm, s.t.|
Frédéric Simonis: Parallel Sub-sampling of Unstructured High Order Data for Visualisation
In order to refine the tetrahedral output mesh of the software SeisSol, the challenge is analysed and strategies are proposed. Being the central problem, Tetrahedral refinement methods are discussed and analysed. With this knowledge, the two higher level problems on how to sub-sample the basic functions and how to refine the mesh are tackled.
Karthikeya Sampa Subbarao: Estimation of Probability Density Functions and Graphical Models Using Regularized Sparse Grid
In this research, we study the application of sparse grid based methods for estimation of probability density functions. In order to improve the performance of density estimation, we device a method to coarsen the grid by identifying and removing, less important ANOVA components from the model. This process also reveals the underlying conditional dependence structure of the dataset's random variables. We employ Markov Chain Monte Carlo sampling for estimating the expected values of sufficient statistics of the model.