Figure: Optimisation landscape of a quantum machine learning algorithm, source: DLR
A (quantum) machine learning algorithm learns by searching for the best model for a given data set. This learning process can be represented in an optimisation landscape: The better the value, the closer the machine learning is to its goal. This optimisation landscape has been derived from a quantum tensor network that detects wing flutter as part of the ELEVATE project.
Quantum computers calculate on the basis of quantum effects and therefore differ fundamentally from conventional computers. The development of modern quantum computers is currently reaching a phase in which the first applications are being implemented in practice and offer initial advantages over classical programs. To support DLR's Quantum Computing Initiative (QCI), the ELEVATE project (Enhanced probLEm solVing with quAntum compuTErs) provides a network to support scientists from all DLR institutes in analysing the feasibility of quantum computer projects. The aim of the project is to enable DLR institutes to work with quantum computers and to establish a strong position for DLR in the field of quantum computing methods.
Supporting DLR institutes in the application of quantum computers
In addition to the Institute for AI Security, quantum computing experts from the Institutes of Software Technology, Quantum Technologies, Space Flight and Astronaut Training, Technical Thermodynamics and Remote Sensing Technology are involved in ELEVATE. Since January 2022, we have been offering further training events, networking meetings and, as part of a short study, an assessment of the feasibility of a topic for the use of quantum computers. The quantum computing experts offer methodological support in four areas
Quantum machine learning applications
Combinatorial optimisation
Quantum simulation
Compilation and hardware
This expertise is taught in lectures, workshops and hackathons so that theoretical knowledge can be put into practice immediately.
Contribution Institute for AI Safety and Security
The Institute for AI Safety and Security contributes its expertise to topics in the field of quantum AI and data encoding. We have conducted pilot studies on the following topics:
Hydrodynamic simulation on quantum computers
Quantum AI for the detection of wing flutter
Navigation of self-driving vehicles with quantum reinforcement learning
Transmission and analysis of aerial images on quantum computers
Anomaly detection in satellite data
Use of tensor networks as quantum AI architectures
Our contribution focuses in particular on the development of transferable design principles between the individual tasks, the implementation of efficient encoding of data and problems on quantum computers and the embedding of quantum models in entire AI environments.
The insights gained in the exploration studies are fed back into the community in workshop formats in order to generate further ideas and find collaborations for follow-up projects.
