The joint project 'AutoQML' was launched in 2022 with the aim of developing solutions that combine quantum computing and machine learning. The aim is to support companies in exploiting the potential of digitalization and remaining competitive. The project is now coming to an end and the partners, consisting of the Fraunhofer Institutes IPA and IAO and seven other companies, are presenting the open source software 'AutoQML'.

Machine learning (ML) already plays a central role in the digitalization strategy of many companies and enables more efficient processes and new business models. However, there is often a lack of specialists, which means that the implementation of ML solutions involves a considerable amount of work. From data acquisition and the selection of suitable algorithms to the optimization of training, this process requires detailed ML expertise.

The automated machine learning (AutoML) approach addresses these challenges and facilitates the use of artificial intelligence. In particular, the selection of ML algorithms is automated so that users require less ML knowledge and can concentrate more on their actual processes.

In this context, quantum computing promises new solutions that significantly improve the AutoML approach. In addition, quantum computing provides the computing power often required for AutoML.

New approach: quantum computing takes machine learning to a new level

The joint AutoQML project built on this innovation and achieved two key objectives: Firstly, the new AutoQML approach was developed, which extends the AutoML principle with novel quantum ML algorithms. Secondly, quantum computing takes the AutoML approach to a new level, as certain problems can be solved more efficiently and sustainably with quantum computing than with conventional algorithms.

The autoQML team at a project meeting in 2023. © Fraunhofer IAO

Under the leadership of the Fraunhofer Institute for Industrial Engineering IAO, the AutoQML open source software developed enables developers to access classical and quantum ML algorithms more easily. The quantum ML components and methods developed were bundled as a toolbox and made available to the development teams. This enables users to use both machine learning and quantum machine learning and to develop automated hybrid overall solutions.

In addition to the Fraunhofer Institute for Manufacturing Engineering and Automation IPA, the companies GFT Integrated Systems, USU, IAV Ingenieursgesellschaft Auto und Verkehr, KEB Automation, Trumpf and Zeppelin also took part in the project. The solutions developed were tested using specific use cases from the automotive and production sectors.

Benchmarking study proves potential of AutoQML

In the final benchmarking study, the project consortium compared its open-source AutoQML software with the best known classical and quantum-based methods. A key result: the automated solutions of the AutoQML software achieve at least equivalent results to the best manually found classical and quantum methods. This gives developers the opportunity to test their own use cases.

The open source software represents a significant step towards a broader application of quantum machine learning in industry, which can sustainably increase the competitiveness and innovative strength of companies.

The progressive market dissemination by the corporate partners promotes the transfer of research-related high technology to a broad industrial environment and significantly strengthens Germany as an industrial location. The scientific findings from the project have been published in several publications. The project was funded over a period of three years by the Federal Ministry of Economics and Climate Protection (BMWK).