Machine Learning meets Quantum Computing

Quantum computing – from theory to reality

The development in quantum computing is progressing rapidly: from a purely theoretical discipline, the first, real-world applications are emerging with universal, programmable quantum computers. While traditional computers work with conventional computing methods, quantum computing exploits quantum effects to perform a computation. Ideal quantum computers promise acceleration in solving specific problems such as mathematical optimization, machine learning, or quantum simulation. However, currently available quantum hardware is typically characterized as "NISQ" for "noisy intermediate-scale quantum." This is because this hardware has only a limited capacity for processing information and is susceptible to errors that can disrupt the actual computational process or readout of the computational result. Nevertheless, it is speculated that in certain applications, even NISQ quantum computers can provide an advantage over conventional computers and algorithms. Although at this stage, there are still many challenges to overcome in developing ideal quantum hardware, it is important to proactively explore potential application areas and become familiar with algorithms and principles of programming quantum computers, automating development processes and extending classical DevOps approaches to the quantum computing domain. 

Our approach

How are we involved with quantum computing technology at Fraunhofer IIS – Department of Localization and Networking?

"Based on our expertise in quantum computing and machine learning, particularly deep reinforcement learning, we are developing both software tools and algorithms to make quantum computing usable for industrial applications in the near future. The interdisciplinary team of experts from the fields of physics, computer science and mathematics provides essential contributions to the quantum computing software stack and creates promising solution approaches that will make quantum computing, especially on NISQ hardware, easily accessible while providing the largest possible computing power," says Dr. Daniel Scherer, Program Manager for Quantum Computing in the Institute's Localization and Networking Division at Fraunhofer IIS.

 

© Fraunhofer IIS

Our quantum computing research areas at a glance:

  • Quantum Algorithms for Machine Learning - We are exploring the use of quantum computers to improve machine learning methods.
  • Quantum Circuit Cutting – We are researching and developing methods for "cutting" quantum circuits. This enables distributed quantum computing on multiple quantum processing units.
  • Quantum Circuit Compilation - We are working on optimized translation of quantum circuits into executable quantum computer instructions.
  • Quantum Error Correction - We are exploring methods for generating hardware-adapted error correction to lower the requirements for error-corrected and fault-tolerant quantum computers.

The Quantum Computing Stack comprises software and hardware necessary to execute quantum algorithms on a quantum computer. In the Localization and Networking Division at Fraunhofer IIS, research and development are focusing on quantum computing middleware: Quantum Circuit Cutting, Compilation and Quantum Error Correction. We also explore quantum algorithms to improve reinforcement learning methods, a special type of machine learning. 

Leveraging machine learning for quantum computing

Machine Learning can contribute to improve the performance of quantum computers and extend capabilities to handle complex, application-specific tasks. Within the scope of our research projects, we focus on the exploration and development of tools and algorithms based on machine learning techniques to enable faster and more efficient computational results using quantum hardware.

To execute quantum algorithms on quantum hardware, corresponding quantum circuits, consisting of a sequence of quantum gates, must be carefully designed and optimized. We are working on developing tools for optimized compilation of quantum circuits ("quantum circuit compilation") using machine learning. This allows  sequences of quantum gates to be rewritten, shortened, and made less error-prone. However, the number of available qubits also plays a crucial role. Here, we are developing software tools for so-called "quantum circuit cutting", which allow us to reduce the number of qubits required to execute a quantum algorithm and ultimately enable distributed quantum computing with multiple quantum processing units. This method contributes to reduce the error-proneness of NISQ hardware and will have a special role on the way to fault-tolerant quantum computers.

The development of error correcting codes for future quantum computers holds further optimization potential. Using machine learning methods, we are developing quantum error correction procedures adapted to quantum hardware. This is also an important step towards reliable quantum computers.

Quantum DevOps tools

DevOps is an approach to software development that emphasizes the combination and integration of development and operations (Dev and Ops). The goal is to accelerate and improve software development and deployment by having development and IT teams work more closely together using common tools and processes. Our software tools described above can be used in a Quantum DevOp toolchain and automate processes for developing and deploying quantum algorithms. DevOps cycles specifically tailored to quantum computing enable continuous delivery of high-quality developments and rapid responsiveness to customer needs and demands.

Reinforcement learning with quantum computers

Quantum computers have the potential to solve problems that are beyond the computing capabilities of classical computers. However, they require a completely different software development approach that differs from conventional programming. In this regard, there are two types of quantum computers: a powerful but not yet available computer that requires universal, error-corrected quantum hardware, and a NISQ computer that already exists but is less powerful and therefore relies on tailored quantum algorithms. Our research focuses on how to work best with both types of computers to achieve good results. To this end, we are developing quantum algorithms to realize reinforcement learning methods on error-corrected hardware, as well as tailored quantum algorithms for NISQ technology based on so-called variational quantum circuits. 

Our offer for your quantum computing project

© Vittaya_25-stock.adobe.com

Fraunhofer IIS is your partner to evaluate the potential of quantum computing-based solutions for complex challenges in your business processes. Thanks to our diverse and interdisciplinary competencies, we develop software tools that can be seamlessly integrated into your quantum DevOps toolchain and automate development processes. Our technology leverages machine learning techniques for the optimized utilization of quantum hardware resources: expand the complexity of your use cases and get a more realistic picture of practical quantum benefits for your applications. We provide consultations and best practices to promote an effective quantum DevOps culture. Let us work together to meet future challenges and achieve your quantum computing goals.

Projekte

 

QACI

 

 

As part of the QACI consortium (Quantum Algorithms for Application, Cloud & Industry), we are researching machine learning-based approaches and methods for optimal use of NISQ hardware through optimized compilation.

 

BayQS

 

 

In the Bavarian Competence Center Quantum Security and Data Science (BayQS), we are researching the use of quantum computers to improve machine learning.

 

QuaST

 

 

In the QuaST research project, we are developing tools to improve the exploitation of quantum computers and paving the way for distributed quantum computing.

 

QLindA

 

 

In the project QLindA we are developing quantum reinforcement learning algorithms for industrial applications 

 

Bench-QC

 

 

In the Munich Quantum Valley lighthouse project Bench-QC, we are exploring benchmarking methods to identify practical quantum advantages.

Further information

ML4POS

Indoor positioning/position estimation with deep learning that replaces the signal processing chain.

Autonomous systems

Automation has become a ubiquitous part of our everyday lives and takes the burden off the user – that is, until a problem occurs. These errors can now be rectified by the machines themselves.

BayQS - Fraunhofer AISEC

The Bavarian Competence Center Quantum Security and Data Science (BayQS) researches relevant software problems in the context of quantum computing.

Online magazine

Series: Quantum Technologies