Quantum meets AI: DLR Institute for AI Safety and Security presents future technologies at ESANN 2025
Scientific session on quantum computers and quantum-inspired methods for machine learning
The intersection of quantum computing and artificial intelligence offers a completely new perspective on the future of machine learning. At the 33rd European Symposium on Artificial Neural Networks (ESANN 2025), the DLR Institute for AI Safety and Security presented its latest research findings on this forward-looking topic.
The scientific session, led by the Institute, focused on innovative approaches to extending and improving classical AI systems through quantum technologies. Topics covered included classical pre-processing for quantum AI, efficient coding strategies, and quantum-classical hybrid models, as well as purely classical machine learning approaches inspired by quantum physics, such as tensor networks.
Four highlight topics demonstrate the potential of quantum methods
Our session focused on four key topics that demonstrate the potential of quantum methods for AI research:
1. The efficient encoding of hyperspectral image data using tensor networks for classification tasks, which is a crucial step in the practical application of quantum computers for image processing.
➡️Encoding hyperspectral data with low-bond dimension quantum tensor networks
2. Tensor networks with normalisation constraints for efficient quantum machine learning using DMRG (Density Matrix Renormalisation Group): an approach that can greatly enhance the computational efficiency of quantum AI systems.
➡️Quantum Tensor Network Learning with DMRG
3. An innovative hybrid quantum annealing approach for predicting excavator prices, developed by Fraunhofer IAO, is an example of the practical application of quantum methods in industry.
➡️Quantum Annealing based Feature Selection
4. The analysis of the trade-off between the expressivity and generalisation capability of quantum kernel methods is fundamental to understanding the theoretical limits and potential of quantum AI.
➡️Expressivity vs. generalisation in quantum kernel methods
'ESANN 2025 provided an ideal platform for scientific exchange with leading figures from the international AI community,' explained the head of the DLR session. 'In particular, the hybrid quantum AI approaches that combine classical methods with quantum technologies were met with great interest.'
Interdisciplinary research for safe AI technologies
Intensive discussions on quantum tensor networks and quantum-inspired methods opened up new research perspectives and created valuable contacts with other experts. These findings will be incorporated directly into the institute's future work.
At the DLR Institute for AI Safety and Security, we develop technologies that enable the safe and reliable use of AI through interdisciplinary collaboration, including with quantum methods. The ESANN findings help us to actively shape the future of AI.
The institute looks forward to advancing its research further and exploring the possibilities of quantum AI with the growing quantum machine learning (QML) community.
Background ESANN
Since its inception in 1993, the European Symposium on Artificial Neural Networks has established itself as a leading event for researchers specialising in the fundamentals and theoretical aspects of artificial neural networks, computational intelligence, and machine learning. The annual conference's combination of lectures and poster sessions once again provided an ideal platform for scientific exchange on current developments in 2025.
Further information on the research of the DLR Institute for AI Safety and Security at the intersection of quantum technology and AI can be found here.