SC²S Colloquium - February 27, 2015

From Sccswiki
Jump to navigation Jump to search
Date: February 27, 2015
Room: 02.07.023
Time: 10:30 am, s.t.
Invited by: Dipl.-Math. Benjamin Uekermann, Univ.-Prof. Dr. Hans-Joachim Bungartz

Alexander Shukaev: A Distributed Communication Model for the Code Coupling Library preCICE

Precise Code Interaction Coupling Environment (preCICE) is a coupling library for partitioned simulations of multiphysics scenarios. Its current release version only offers a centralized communication model (CCM) between processes of coupled solvers, that is they communicate in a client/server manner. Performance testing on massively parallel systems has confirmed that this approach results in a data throughput bottleneck (on servers), which leads to a dramatic loss of scalability. The potential solution to this problem is to develop a distributed communication model (DCM) that would be based on a parallel process-to-process communication approach. In this talk, I will present the corresponding software engineering challenges, their goals and solutions.

Viacheslav Mikerov: Using a Fixed Grid Finite Difference Flow Solver for Partitioned Fluid-Structure Interaction

The research is dedicated to the partitioned Fluid-Structure Interaction (FSI) approach with a minimal invasive software design using PreCICE (Precise Code Interaction Coupling Environment) library. A specific goal of this work is to present such a fluid flow solver on a fixed grid using a finite different discretization which can interact with different structure solvers. One of the challenges here is to handle moving or deforming bodies with complex surface geometry embedded in a fluid flow.

Prof. Dr. Massimo Fornasier: Numerical methods for sparse recovery

In this talk we review numerical methods for performing optimizations with linear model constraints and additional sparsity conditions to solutions, i.e., we expect solutions to be represented as sparse vectors with respect to a prescribed basis. In the first part of the talk we illustrate the theory of compressed sensing with emphasis on computational aspects. We sketch the analysis of the iteratively re-weighted least squares method, the iterative hard-thresholding, and more general primal-dual algorithms. In the second part, starting from the analysis of iterative soft-thresholding, we illustrate several numerical methods for addressing sparse optimizations in Hilbert spaces. The third part is concerned with numerical techniques, based on domain decomposition methods, for dimensionality reduction in large scale computing. We eventually mention extensions to low-rank matrix recovery and its numerical challenges.