Adversarial and domain shift robustness for speaker recognition (COMFORT)

Third Party Funds Group - Sub project


Acronym: COMFORT

Start date : 01.10.2024

End date : 30.09.2027


Overall project details

Overall project

Komprimierungsmethoden für Robustheit und Transfer

Project details

Short description

Modern machine learning methods such as deep neural networks or generative models have the potential to revolutionize science, industry, and our society. To deploy machine learning more broadly and sustainably, it is crucial to advance the development of compact and robust models that are suitable for a wide range of data and applications. In the COMFORT project, we strive for breakthroughs in the development of compact, flexible, and robust machine learning models for diverse applications in the areas of image, audio, and network data. To achieve this, our application-oriented research program will advance the mathematical understanding of machine learning at the interface of effectiveness and robustness. The project goals are to develop a theoretical understanding of the relationship between robustness and efficiency, a framework for novel compression techniques, efficient training and evaluation algorithms, generalizability to various application scenarios, and industrial transfer. To achieve this, COMFORT has assembled a consortium consisting of researchers from the Center for Artificial Intelligence and Data Science (CAIDAS) at the University of Würzburg, the German Electron Synchrotron (DESY), the Technical University of Munich, the University of Hamburg, the Friedrich-Alexander-University Erlangen-Nuremberg, and AudioLabs. This consortium bridges the gap between mathematics, computer science, and innovative applications. Pruna AI, a Munich-based technology startup specializing in environmentally friendly and efficient artificial intelligence, serves as an industrial partner. The ambitious work program of COMFORT combines efficiency, resilience, and generalizability, aiming to create leading locations in Germany for sustainable, reliable, and versatile machine learning.

Involved:

Contributing FAU Organisations:

Funding Source