SynData develops advanced methods for generating synthetic research data using AI while ensuring their quality and making the creation process transparent in line with "Explainable AI" principles. The project strengthens expertise in research data management, interdisciplinary data utilization, and the responsible use of AI-generated data.
In addition to "in vivo" data collected from real-world applications, countless laboratory experiments generate experimental "in vitro" data. With the advancement of machine learning-trained models, it is now possible to produce synthetic "in silico" data. This type of data generation enables the creation of new datasets and allows for post hoc filling of gaps in existing datasets.
As part of DataNord, the SynData project explores these innovative approaches to generating research data. It focuses on modeling the quality of synthetic datasets, expanding existing models of dataset structures, and incorporating explainability components into black-box methods for synthetic data generation. These components, aligned with the principles of "Explainable AI", aim to help users of synthetic datasets understand how specific results are derived.
SynData contributes to making the application of AI models for data generation more user-friendly and transparent – addressing a critical challenge for the data science community. By tackling current data science issues that impact all scientific disciplines, the project strengthens competencies in handling these emerging technologies. Through the development of transferable skills and methods, SynData creates broad value and supports the sustainable advancement of data science.
Additionally, SynData enhances data competencies in DataNord's key focus areas of "Material and Engineering Sciences" and "Social Sciences".
Team Members
University of Bremen: Prof. Dr. Rainer Malaka, Dr. Robert Porzel, Lennard Scheurer
Fraunhofer IFAM: Prof. Dr. Bernd Mayer, Dr. Katharina Koschek, Dr. Vinicius Carrillo Beber
