About Us

The research group Theory and Applications of Algorithms (TAA) studies combinatorial as well as numerical algorithmic challenges both from a theoretical as well as from an application-oriented viewpoint.
» Prof. Monika Henzinger and her group focus on combinatorial questions, » Prof. Wilfried Gansterer and his group concentrate on challenges in numerical algorithms. The two groups interact and cooperate in several areas, such as data mining and machine learning as well as graph theoretical problems arising in distributed computing.

In the field of combinatorial algorithms we currently focus on two aspects, namely graph algorithms and combinatorial optimization and algorithmic game theory.
 

Graph algorithms and combinatorial optimization

Graphs are used to model relationships between items in a wide variety of applications. One such application is computer aided verification, where the system to be verified is modeled as a graph and properties of this graph are used to determine the correctness of the system. We study graph algorithmic questions arising in computer aided verification as well as combinatorial optimization problems on graphs that arise in other applications such as computational biology and computer vision.
 

In the field of numerical algorithms, » Prof. Wilfried Gansterer’s Research Lab Computational Technologies and Applications (RL CTA) works on parallel and distributed algorithms for numerical matrix computations, on mixed precision algorithms, on fault tolerant algorithms, and on related application problems arising in computational science, in data mining and machine learning, and in internet security. The focus is on high perfor- mance aspects as well as on algorithmic adaptation to modern hardware platforms.

Parallel and distributed matrix computations

Designers of state-of-the-art computer systems continuously increase hardware concurrency at various levels. This widens the gap between the rapidly growing theoretical hardware peak performance and the performance achieved by existing algorithms and software. In order to narrow this gap and to improve the often poor efficiency of numerical computations, we investigate new algorithmic strategies which exploit concurrency at various levels. Our focus is on prototypical matrix computations for hardware platforms ranging from multi and many-core systems to loosely coupled sensor networks.
 

Mixed precision algorithms

Modern heterogeneous computing platforms often contain components where low precision computations can be performed much more efficiently than high precision computations (e.g., GPUs or FPGAs). We study algorithmic concepts for utilizing several precision levels and for controlled trading of accuracy against performance.
The research of »Prof. Gansterer is supported by the FWF, the FFG and by industrial partners.