NHR PerfLab Seminar on July 11: Towards robust and efficient AI at scale (online)

Symbolbild zum Artikel. Der Link öffnet das Bild in einer großen Anzeige.

Title: Towards robust and efficient AI at scale

Speaker: Dr. Charlotte Debus, Junior Research Group Leader “Robust and Efficient AI” at the Steinbuch Centre for Computing, Karlsruhe Institute of Technology (KIT)

Date and time: Tuesday, July 11, 2023, 2:00-3:00 p.m. CEST

Slides

Abstract:

Next to everyday life applications like smart phones, autonomous driving or voice assistants, AI methods have also revolutionized data analysis, system monitoring and control optimization in scientific research and engineering. In these domains, data is often acquired over time, adding an extra dimension and as such more complexity to the problem. The introduction of transformer architectures has opened up new ways of accurately predicting behaviors in such dynamic systems.

However, the sheer size of transformer models puts AI research in front of new challenges, pushing towards large scale models that run on multiple accelerators and supercomputers. The increased demand in compute resources comes at the price of growing energy consumption, which now raises the question regarding sustainability and environmental friendliness of AI applications.

In the talk, we will look at a few examples of scientific applications of AI-based time series forecasting as well as their scalability and energy efficiency, and what we can do to balance these apparently competing trends of Scalable AI and Green AI.

Short Bio:

Charlotte DebusCharlotte Debus is an applied AI researcher and junior group leader at the Steinbuch Center for Computing (SCC) at the Karlsruhe Institute of Technology (KIT). She studied physics at the University of Heidelberg and completed her PhD on the topic of machine learning methods for medical image analysis in 2016. After spending 2 years as a research associate at the German Aerospace Center (DLR), she moved to KIT in 2020, where she joined the Helmholtz AI consulting team and developed AI methods for scientific applications in the field of energy research. In 2022 she successfully obtained a grant by the German Federal Ministry of Education and Research for a junior research group. Her research focuses on developing robust and efficient AI methods at scale, especially for time-series forecasting problems.


For a list of past and upcoming NHR PerfLab seminar events, see: https://hpc.fau.de/research/nhr-perflab-seminar-series/