Industry and business cite two problem areas across sectors in connection with Big Data applications and machine learning (ML): On the one hand, the lack of sufficiently many and, above all, well-trained data analysts is emphasized, and on the other hand, there is usually not enough data available for technical applications, for example, to train large neural networks (NN) in a stable manner via Deep Learning (DL) approaches.
This project therefore aims to address such a prototypical digital pathology problem and to analyze and methodically implement mathematically sound procedures for data augmentation via neural networks/Deep Learning. The AG Technomathematik of the University of Bremen coordinates this joint project and is involved on the scientific side with the subproject Invertible Network Architectures for Data Augmentation. Here, invertible network architectures (i-RevNet, learned Mixup) are analyzed for data augmentation and used to expand the database in a histopathological application.
In addition, the measures accompanying the program are coordinated within the framework of the Mathematics for Innovation funding priority. Among other things, Innovation Labs and Challenge Workshops are planned here to accelerate the transfer of project results from all funded collaborations to industry, society and the public.
More information on the program's accompanying activities can be found here: https://math4innovation.de