GPU Power for Deep Learning Applications
Over the last couple of weeks, the GPU server of the study programme Data Science and Business Analytics was optimised for applications in the field of artificial intelligence. This means that our students now have a high-performance environment for various Deep Learning applications at their disposal.
With the support of Alexander Buchelt who is currently completing his mandatory internship at the St. Pölten UAS, our Data Science Core Team made up of Alexander Adrowitzer and Torsten Priebe carried out extensive tests with complex neural networks of large data volumes on the server.
Advantages for Lecturers and Students
The parallel use of all eight NVIDIA GeForce 1080 Ti GPUs yields enormous advantages in terms of speed, thus opening up entirely new fields of application in teaching and research. For example, large amounts of images can be classified effortlessly in order to give the students an understanding of the possibilities and risks of face recognition.
Starting from next semester, the powerful hardware with two ten-core CPUs is available to all students of the bachelor degree programme Data Science. Students specialising in healthcare can use the optimised hardware to develop their own prognosis models, for instance regarding the spread of pandemics. When it comes to production, large data streams can be analysed in order to predict potential failures of machines in a timely manner.
Application in the Data Science Bootcamp
Within the framework of the Data Science Bootcamp which takes place this autumn with grants from the Austrian Research Promotion Agency (FFG), we will use the GPU cluster to present possible applications in various disciplines – from the evaluation of mobile communications data to the provision of weather forecasts. The Digital Pro Bootcamp is a nine-week further training for staff members of eight small and medium-sized businesses at the St. Pölten UAS in the area of data analysis with a focus on data engineering, data visualisation and artificial intelligence.
Next Expansion Step Planned
The next upgrade of the data science infrastructure is already under way: the server is being expanded into a cluster in order to provide the students with a distributed Big Data architecture based on Hadoop and Spark for learning and training purposes.