Tübingen Machine Learning Cloud

• Term: Funded in 2019
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 419240949

Over the course of only few years, machine learning has quickly evolved from a niche into a key technology within Computer Science. Together with it's local partners, the University of Tübingen hosts nationally outstanding expertise in this area. Tübingen is also among the leading locations internationally in the machine learning community. A number of recent new hires is helping to fortify this competitive position. The funds for the "Tübingen ML Cloud" requested in this proposal aim to provide modern and ambitious IT infrastructure for the research groups working in this area.Machine learning algorithms operate on large datasets and utilise massively parallel computations. Both of these aspects pose new challenges on computing infrastructure that are not yet served well by classic high performance computing centres. Graphical processing units (GPUs) play a central role in this area. High demands are also placed on storage and networking, to facilitate fast and flexible handling of large volumes of data. In contrast to classic high performance computing, parallelism in machine learning is primarily (though not solely) handled on the individual machine. The technology initiative combines central funds of the President's office of the University with startup-funding of recently established new Chairs to achieve a lasting and significant improvement of the IT-Infrastructure. It comprises the acquisition of central computing workstations with high-performance GPU cards, significant storage space and high performance networking infrastructure. The computing resources will be made available to users in the context of a "Cloud" metaphor (i.e. through individual virtual machines). This will reduce or remove usability barriers for users and allow flexible use by a range of users.

• DFG Programme: Major Research Instrumentation
• Major Instrumentation: Tübingen Machine Learning-Cloud
• Instrumentation Group: 7000 Datenverarbeitungsanlagen, zentrale Rechenanlagen
• Applicant Institution: Eberhard Karls Universität Tübingen
• Leader: Professor Dr. Philipp Hennig