Promotion of efficient machine learning models

September 18, 2023

The performance of artificial intelligence (AI) has increased significantly in recent years. This is primarily due to advances in the field of learning systems. Such deep-learning models are also beginning to accelerate scientific progress.

Current machine learning models are usually specialized for a specific task. In the future, however, flexible models capable of solving a variety of tasks and handling different types of data will gain in importance. At the same time, the requirements for the resilience of learning systems in practical use are also growing.

As the complexity of models increases, working groups will increasingly need to collaborate across disciplines and institutions to share their expertise as well as their data and computing resources. To this end, there is a need for widely available challenging benchmark datasets and environments to compare the flexibility, robustness, and efficiency, as well as bias and fairness, of different systems.

Funded are interdisciplinary AI projects to develop new model architectures and learning algorithms to improve the flexibility, resilience, and efficiency of learning systems or the efficiency of simulation models.

To avoid overlap with other funding areas, no projects will be funded here that aim to develop AI hardware, neuromorphic computing, or the use of AI in medicine, human resources, marketing or customer care, IT security, predictive maintenance, civil security, or robotic systems for care. Furthermore, application-driven or interdisciplinary projects must generate added value for AI research, which must still be demonstrated in concrete use cases within the project term.

The grants are awarded by way of project funding as a non-repayable grant.

The BMBF has currently commissioned the following project management organization to handle the funding measure: Deutsches Zentrum für Luft- und Raumfahrt e. V., DLR Projektträger (PT-DWS/SIS),  Sachsendamm 61, 10829 Berlin. Contact persons are Dr. Ulrike Wunram and Mr. Lars Mehwald, e-mail: Contact by Mail .

The application procedure has two stages. In the first stage of the procedure, by January 12, 2024 project outlines are to be submitted in electronic form. In the second stage of the procedure, the authors of the positively evaluated project outlines will be invited to submit a formal funding application.

>> Further information