Advanced AI and HPC - Quantum Computing
The team Quantum Computing brings together DFKI researchers with a strong scientific background in physics, math, and computer science with academic research teams from the University of Bremen. Our research focuses on various areas of Quantum Computing and Quantum Machine Learning, namely:
- Quantum-assisted machine learning
- Quantum-inspired machine learning
- Quantum error correction and error mitigation
- Quantum Optimal Control
- Distributed Quantum Computing
The first part of our work is transferring research results towards prototypical usage in industry. Secondly, we are improving quantum computing technology to make it ready for industrial usage.
For the first part, in the area of quantum-assisted machine learning, we evaluate how quantum computing resources can be made useful for supervised learning, unsupervised learning, and reinforcement learning. For quantum-inspired machine learning, we rely on tensor network techniques that have been established in physics. The explainability one can gain here allows our industry partners to use Artificial Intelligence in regulated areas by deploying a model that can inform them why a decision was or was not made. Furthermore, quantum-inspired techniques are also used to develop variational quantum solvers for Computational Fluid Dynamics. This helps saving resources in aerospace and mobility simulations as well as transferring towards renewable energies and ecologically friendly designs.
Improving quantum computing technology in the current Noisy Intermediate-Scale Quantum (NISQ) era requires work towards the correction of errors as well as their mitigation. We work on benchmarks for quantum hardware using error correction codes and also develop novel error mitigation techniques suited for use in quantum machine learning. In the area of quantum optimal control, we use techniques such as optimization methods that are established in optimal control for robotics and transfer these to the control of quantum hardware through laser pulses. Finally, for distributed quantum computing, we investigate methods to make efficient use of several quantum hardware resources that are connected through classical and/or quantum channels.
Projects:
Publications:
- Quantum Deep Reinforcement Learning for Robot Navigation Tasks: https://ieeexplore.ieee.org/document/10568148
- Learning Capability of Parameterized Quantum Circuits: https://arxiv.org/abs/2209.10345
Teamlead: Dr. rer. nat. Gunnar Schönhoff
Deputy: Dr. Elie Mounzer