The Chair for Operating Systems (Univ.-Prof. Dr. habil. Th. Bemmerl) announces the following master thesis on the topic Parallel Computing / Multicore Architectures:
Data Mining Algorithms on Multicore Architectures
(Hoch-paralleles Data Mining auf Multicore Rechnern)
in cooperation with the Chair of Computer Science 9 (Univ.-Prof. Dr. rer. nat. Thomas Seidl) and IBM Deutschland Research & Development
©IBM
Task:
Data Mining allows us to find useful patterns in large information
collections. However, most Data Mining algorithms are computationally very
demanding. Although many improvements in efficiency and accuracy have been
achieved, pure software solutions have reached their limits. A promising
way to overcome these limitations is to use specialized parallel hardware
such as the Cell processor or Graphical Processing Units (GPU) found on
many current graphic cards.
This master thesis will explore the potential of Cell and GPU technology
for Data Mining. More specifically, a hybrid system for clustering data
streams will be developed, that is on the one hand fast enough to allow for
interactive Data Mining, but on the other hand can process vast amounts of
data. The performance of a Cell, GPU and CPU based implementation will be
compared for several real-life mining workloads.
This master thesis will be conducted at IBM Deutschland Research & Development in Böblingen, Germany.
Requirement Profile:
- Bachelor in Electrical Engineering and Information Technology or in Computer Science
- basic knowledge in parallel computing
- good knowledge in C and C++
- ability to work both autonomously and creatively
Contact:
Ruben Niederhagen
Chair for Operating Systems (LfBS), RWTH Aachen University
Modulbau Kopernikusstraße, 52074 Aachen
Tel.: +49 241 80 27699
E-Mail: ruben@lfbs.rwth-aachen.de
| 
