Munich Quantum Valley
A quantum leap for machine learning
Fraunhofer IKS is carrying out research into safe software applications for quantum computers. The work is part of the new Bavarian initiative Munich Quantum Valley.
Quantum computing has the potential to bring huge and sustainable change to a large number of industries and pave the way for numerous new applications. These include, for example, simulations of new materials for more efficient solar cells and batteries, solutions for complex optimization problems in the logistics, finance and insurance industries, and data-intensive applications for artificial intelligence and cybersecurity.
Find out more about Munich Quantum Valley in the press release of the Fraunhofer-Gesellschaft:
The “Munich Quantum Valley” was presented in Munich yesterday to promote the development of quantum science and technology as part of the Bavarian quantum initiative. In addition to the Bavarian Academy of Sciences and Humanities, Ludwig-Maximilians-Universität München (LMU Munich), the Max Planck Society and the Technical University of Munich, the Fraunhofer-Gesellschaft is also a partner of the initiative. Munich Quantum Valley is aiming to strengthen research, development and vocational training in the field of quantum science and technology.
The Fraunhofer-Gesellschaft will be taking part, both at the hardware level with the quantum computers that are currently in development with IBM, and at the software level through the Bavarian Competence Center for Quantum Security and Data Science (BayQS). Within BayQS, the three Fraunhofer institutes AISEC, IKS and IIS are pursuing the goal of playing a crucial part in the design of quantum computing, for example by developing quantum algorithms.
Small particles, big effect
A “quantum” describes the smallest possible physical size that exists. It cannot be divided, but only created or destroyed in its entirety. Although there has been research into quanta since the 1960s, physicists still do not fully understand them because the laws of quantum physics contradict our day-to-day experiences.
Safety expertise for quantum computing through Fraunhofer IKS
As part of the development of software applications for quantum computing, the Fraunhofer Institute for Cognitive Systems IKS is harnessing its safety expertise to focus on the research field of “Reliable and Robust Quantum Computing”.
Though quantum computing offers many possibilities for innovative solutions to problems, it can only provide real added value if the calculations are reliable and certain. As well as the computing power itself, the robustness of quantum computing is also a fundamental requirement for successful implementation in practice. To solve this challenge and design software applications for quantum computers reliably, Fraunhofer IKS is primarily researching two topics: quantum computing-supported verification of neural networks, and reliable quantum computing-supported AI for medical diagnostics.
Proof of quality for neural networks using quantum computing
The use of neural networks in safety-critical environments is highly challenging, as even small changes in an input, such as an image, can cause the calculations to come up with different results. In order to make a reliable statement about the quality of the calculations of the neural network and thereby design it more safely, particular algorithms are used to provide proof of quality — for example, to establish that a three percent change in specific defined characteristics does not change the result of a calculation made by a particular neural network. However, these verifications of quality currently only work for relatively simple networks, as the calculation time required to provide the proof increases enormously with the size and complexity of the neural network. Fraunhofer IKS is therefore researching how to improve the optimization and security of autonomous, networked systems using quantum computing.
Superposition instead of binary code
Quantum computers are made up of qubits. While bits only have two possible states, 1 and 0, qubits can be in any state in between — this is called superposition. Qubits are many times more powerful than normal bits, but they are also more sensitive. To be as stable as possible, they need to work in an environment close to absolute zero.
Reliability for networked medical diagnostics
As part of the second research topic, Fraunhofer IKS is collaborating with the Neurosurgery and Radiology Clinics at LMU Munich to investigate how quantum computing and artificial intelligence can be used for intelligent diagnostics and in healthcare. The requirements here in terms of the reliability and transparency of diagnoses made by complex, networked systems are high. The focus of the research project is on the screening and follow-up diagnosis of brain tumors. Specifically, it is researching to what extent the use of AI methods allows a better analysis of the raw image data to be produced and to what extent the interaction of different areas of the brain can be analyzed and simulated. The researchers are also investigating how medical decision-making processes can be improved by means of an integrated approach to image data and other data. The use of quantum computers could lead to a breakthrough here.
Quantum computing: From research to industry
As the research projects show, there are still many open questions and challenges surrounding the use of quantum computing in industry. The recently launched initiatives for the development of quantum science and technology, in which the Fraunhofer-Gesellschaft and Fraunhofer institutes such as Fraunhofer IKS are involved, are working to answer these questions and solve the existing challenges in order to design the use of quantum computing safely and reliably for a diverse range of industrial applications.
What has emerged already is that, because of the complexity and high demands of the hardware, quantum computers will not replace conventional computers, but supplement them. Quantum computers will help us to answer very concrete and specific questions that are beyond the limits of today’s supercomputers. For broader use in practice, including in AI, however, traditional computers will continue to be more relevant.