Advanced Topics in High Performance Computing, Data Management and Analytics

  • type: Seminar (S)
  • chair: KIT-Fakultäten - KIT-Fakultät für Informatik - Institut für Telematik - ITM Streit
  • semester: WS 21/22
  • lecturer: Prof. Dr. Achim Streit
    Anis Farshian Abbasi
    Nikolai Pfisterer
  • sws: 3
  • lv-no.: 2400117
  • information: Online
Content

Artificial Intelligence, Big Data, Supercomputing, High Performance Data Analytics, Machine Learning, Exabytes, GPUs, Exaflops, Data-Intensive Computing, FAIR-Data, Quantum Computing - these are all modern terms and topics in the fields of High Performance Computing (HPC), Data Management and Data Analytics.

HPC computing systems can be used to efficiently simulate digital twins of natural and man-made things, processes, and phenomena. Artificial Intelligence and Machine Learning have only become efficiently possible due to the incredible computing power of modern high performance computers and GPUs (and in the future quantum computers). On the one hand, the explosive growth of big data poses an enormous challenge for its performant and sustainable handling, but at the same time it offers the opportunity to transform data into new knowledge and action using computation-hungry analysis techniques.

In the seminar, selected topics of high-performance computing, data management and data analysis will be covered. Keywords are e.g. MPI, OpenMP, CUDA, OpenCL, OpenACC, tools for analyzing efficiency and scalability, parallel and distributed file systems, on-demand file systems, high-speed transfer technologies, scheduling, metadata, data repositories, PyTorch, time series analysis, analysis of Big Data streams, optimization of neural network training, Uncertainty Quantification or graph-based neural networks.

Therefore, efficient methods, tools, concepts, technologies, and approaches will be presented, compared, evaluated, and discussed in the seminar. If required, the elaboration, presentations, and discussions can also be written or held in English. The final list of topics will be presented at the beginning of the seminar.

The first event of the seminar, in which, among other things, the topics will be briefly presented and finally assigned, will take place on Monday 25.10.2021 at 16:00 - online/virtual in MS Teams.

The seminar presentations will be held towards the end of the lecture period in a few block sessions.

In order to be able to estimate the number of participants as well as for the topic allocation procedure, we ask interested students to register for the seminar in advance in the ILIAS system. Topics will be assigned at the first meeting according to the FCFS principle (first-come-first-serve). If several students are interested in the same topic, the order of ILIAS registration will decide.

Registered persons must be present at the first meeting, otherwise no topic can be assigned.

In case of missed ILIAS registration there is still the possibility to get a topic at the first appointment - depending on demand.

Language of instructionGerman/English