The Quantum AI research focus investigates the practical use of quantum algorithms and quantum methods.
Quantum computing is considered one of the next steps in the development of even more powerful supercomputers. Research institutes and technology companies around the world are investing in various types of quantum computing. Northwestern Switzerland also has a strong quantum computing ecosystem, where work is being done on quantum computing and its practical applications.
What contribution do we make to this?
The Quantum AI research focus investigates whether, how and which economic decision-making and optimisation problems can be better solved with quantum computing. To do this, we use GPU-accelerated simulators and various quantum platforms such as annealers, ion traps, superconducting systems and neutral atom systems.
How do we do that?
We formulate decision-making and optimisation problems in such a way that they can be run on both classical computers and quantum simulators and quantum computers. We then compare, for example, solution accuracy, solution speed and scalability in order to draw conclusions about the added value of quantum computing and develop new methods. The findings are fed back into the development work of our research partners and help companies to better assess when and how quantum computing will be useful for them.
Why is this important?
Political, research and technology partners receive guidance on which decision-making and optimisation problems quantum computing has potential for. Our work thus contributes to building quantum expertise in north-western Switzerland, with a clear focus on practical application and as a complement to cutting-edge research in the region.
Our research explains using an example
The knapsack problem on quantum computers
To solve an optimisation problem such as the knapsack problem on a quantum computer, it is translated into a quadratic energy function. Each possible solution corresponds to a point in a landscape of mountains and valleys, with the optimal solution lying in the deepest valley. The rules (e.g., pack each item only once) shape this landscape by being built directly into the objective function, rather than as separate constraints as in classical methods.
Previous approaches used very high weightings to rule out rule violations. However, this distorts the energy landscape and leads to less favourable results. Our experiments show that significantly lower weightings are often sufficient to obtain stable results. This creates a practical basis for research and provides users with concrete guidance on how to make better use of quantum computers.
Publications
