.digital blog

Deep learning and dataLAB – a success story

Growing volumes of data, new questions in research and teaching – all of this requires an increasing bank of hardware and software resources, appropriate design and administration. The TU.it dataLAB and the foundation of the DL community has created a close network between researchers, teaching staff and TU-it (HPC group).

[Translate to English:] Deep Learning, dataLAB

[Translate to English:]

[Translate to English:]

[Translate to English:]

Stronger together

In principle, any institute or research group that has the financial means and technical expertise could build an appropriate computing cluster. However, a more efficient way of doing this is to contact the High Performance Computing Group (HPC) and the TU.it, opens an external URL in a new window dataLAB and take advantage of their services, since they have the relevant infrastructure to implement innovative projects in areas like big data, data analysis or deep learning. The team consists of colleagues with a scientific background or IT training, which gives them a good understanding of their users' needs, to which the available products and applications are adapted. Both sides benefit from this: the HPC group learns from the feedback while the advantage for users is an attractive overall package created by pooling resources. They can access a much larger cluster that is constantly evolving without needing their own IT staff. As TU employees, they can also attend courses and training sessions at any time offered by the Vienna Scientific Cluster (VSC), which more recently have included courses on DL and GPU programming.

Networking without limits – cross-faculty collaboration

From this year, the dataLAB team has also been holding regular networking meetings, where scientists from different institutes and faculties come together to facilitate exchanges of experiences and to promote cooperation. The first two network meetings were well attended. They included talks from the HPC team about existing and planned hardware and exciting motivational talks by users about current projects. A follow-up is planned for the beginning of 2020.

One application on the cluster is the Data Science master's programme. Not only does this teach about DL and data science tools, reflection on ethical issues and equal opportunities are also central themes throughout the course, as Florina Piroi reported. This is very important because data bias could result in false or even dangerous conclusions, take for example justice or medicine. Jiří Hladůvka spoke about his course, which uses a tool (Jupyter), in which students and teachers can issue programming tasks, feedback and assessments. Even integration into an assessment platform (ngrader) is seamless. In general, both teachers and students appreciate that the same version of the software is available for everyone on the cluster without the need for individuals to have to install it themselves.

Curious? Get in touch with the dataLAB team if you are interested in using our cluster or if you would like to attend the next DL and dataLAB community networking meeting. 

Further information:
dataLAB: https://it.tuwien.ac.at/datalab, opens an external URL in a new window

Contact for DL und dataLAB Community networking meetings:
Irene Reichl (Irene.Reichl@tuwien.ac.at)

VSC course website: http://vsc.ac.at/training/, opens an external URL in a new window