5th Symposium AI4Life: Interdisciplinary AI research at the HSWT
Around 200 researchers, students and experts from science and industry gathered information online at the HSWT on developments and possible applications of AI in the life sciences.
This specialist symposium was held at HSWT for the fifth time and was organised by Prof. Dr Claudia Brand, Prof. Dr Patrick Noack and Dr Florian Haselbeck. The range of topics spanned from novel methods of machine learning and the use of AI-based sensor technology to applications in environmental and climate research. The keynote speech was given by Prof. Dr Sepp Hochreiter, Johannes Keppler University Linz, one of the leading minds in AI research.
Transformer models for graphene structures
In his presentation „From language modelling to graph modelling: Transformers are Scalable Graph Generators“, Dexiong Chen highlighted a new framework for graph generation based on transformer architectures. By combining language modelling and graph modelling, graphs are converted into sequences like words and generated using reversible random walk sampling processes.
Condition monitoring of machines with AI-supported vibration sensor technology
Christoph Behr, HSWT and Tobias Kolbe, agricomp GmbH, spoke on „Development and Validation of a Vibration Sensor for State Recognition of CHP Machinery“ and presented a vibration sensor for monitoring the condition of combined heat and power plants. The system uses acceleration sensors on the engine block and enables reliable detection of different fault types and states and is the basis for more efficient maintenance strategies and optimised operating plans.
Recognising heat stress in bumblebees - with AI and acoustics
Sebastian Burkhart and Finja Schaumann, HSWT, presented the interdisciplinary project „Identifying Heat Stress in Bumblebee Colonies Based on Audio Recordings“. The aim is to recognise heat stress in bumblebee colonies based on audio recordings. The results show the potential of scalable monitoring systems for pollinator populations and are an important contribution to the protection of these ecologically important insects.
Analysing animal behaviour with dual-stream AI
Ehsan Yaghoubie, HSWT, presented „Cross-modal uncertainty based fusion for animal behavior analysis using audio“, a model for AI-supported analysis of animal behaviour based on acoustic signals. The „dual-stream model“ combines frequency features and time-based raw audio signals of animal sounds and recognises emotions, behaviour and well-being and can be used for the assessment of animal welfare.
AI for climate protection: EcoMapper generates climate-sensitive satellite images
In his presentation „EcoMapper: Generative Modelling for Climate-Aware Satellite Imagery“, Malte von Bloh presents a generative model that uses satellite images for climate-related applications with the help of stable diffusion methods. This enables realistic, climate-sensitive satellite images with a high resolution, which can close previous data gaps due to cloud cover, for example. This is an advance in generative earth observation and opens up new applications in climate prediction, adaptation strategy and geospatial analysis.
Outlook for future research
The symposium demonstrated the wide range of applications of artificial intelligence in the life sciences - from microorganisms to global climate analysis. The lively discussions and interdisciplinary approaches showed that the combination of AI, life sciences and technology continues to offer enormous potential for sustainable research and innovation at the HSWT.
