From 12.–17.11.2023, ScaDS.AI Dresden/Leipzig took part in Supercomputing 2023 (SC 23) in Denver, Colorado. The conference is the largest and most important platform of the international High Performance Computing (HPC) community, where scientific research results are presented as well as technological innovations and new products. We were happy to present our work at a joined booth (#1553) with our colleagues from Center for Information Services and High Performance Computing (ZIH) and Center for Interdisciplinary Digital Sciences (CIDS). Prof. Dr. Wolfgang E. Nagel, Dr. René Jäkel, Dr. Siavash Ghiasvand, Lena Jurkschat, Neringa Jurenaite and Elias Werner from TUD Dresden University of Technology were the attendees from our center.

Program

ScaDS.AI Dresden/Leipzig used this highly important international platform to present itself as a german competence center for research on Artificial Intelligence, Big Data and Data Science.

Sofa Talks

• Interactive HPC with Jupyter
  with Elias Werner on 14.11.2023, 11 a.m. – 12:15 p.m.

• Efficient Coupling of Highly Parallel CFD Simulations with Machine Learning on Heterogeneous Architectures
  with Elias Werner and Fabian Orland on 14.11.2023, 12:30 p.m. – 1:45 p.m.

• Failure Prediction on HPC
  with Dr. Siavash Ghiasvand on 14.11.2023, 2 p.m. – 3:45 p.m.

• OpenGPT-X Challenges in Training Large Language Models (LLM)
  with Lena Jurkschat on 14.11.2023, 3:30 p.m. – 4:45 p.m.

Demos

Our Living Lab also presented a variety of research demonstrators to the interested audience at Supercomputing 2023:

asanAI

With asanAI, small sequential models can be created and trained in any browser without the need for installation on a local computer. Users can import their own data, including images, webcam data, CSV files, or arbitrary tensor data. Furthermore, the (trained) models can be exported directly to Python code, Node JS code, and HTML code. You can use asanAI yourself here. To ensure an optimal start into using the application, we also provide a playlist with video tutorials on YouTube.

MultiCut

MultiCut is a gamification of the multicut algorithm, which can be used for the image segmentation process in computer vision. Users are challenged to find the best possible multicut in levels with different degrees of difficulty from a simple drawing to detailed scientific images.

Content Based Image Retrieval (CBIR)

CBIR takes an image as the input, and then searches a large database of historical images to find similar images taken from various angles. This demo generates a unified representation of each image via multiple convolutional layers. The resulting representation is used to search a large database of images.

Bridging between Data Science and Performance Analysis: Tracing of Jupyter Notebooks

Bridging between Data Science and Performance Analysis: Tracing of Jupyter Notebooks enables the analysis of the performance of Python applications and Data Science workflows in Jupyter. Score-P is used here as a measurement infrastructure. The recorded performance data can be visualized with Vampir. This is realized by implementing a Jupyter kernel.

Supercomputing 2023

The International Conference for High Performance Computing, Networking, Storage and Analysis is a yearly conference. In 35 active years, the conference has brought together many international experts on High Performance Computing. The conference location changes annually; however, Supercomputing 2023 was held in Denver, Colorado. Find out more about SC23 on the official conference website. ScaDS.AI Dresden/Leipzig and ZIH have participated in the conference for many years. Here, you can read more about our participation in Supercomputing 2022.