Technische Universität Dresden
The Technische Universität Dresden (TUD) is one of the leading and most dynamic universities in Germany and was identified by the German government as a ‘University of Excellence’. The “Chair of Stochastic Models / Reliability, Asymptotics”, chair holder Prof. Dr. Zoltán Sasvári, covers the fields stochastics, mathematical statistics, and their applications.
The Technische Universität Dresden has about 34,000 students and over 8,300 employees, 560 professors among them. As a full-curriculum university with 14 faculties in five schools, it offers a broad variety of 129 degree programmes and covers a wide research spectrum. Interdisciplinary cooperation among various fields is a strength of the TUD, whose researchers also benefit from collaborations with the region’s numerous science institutions – including Fraunhofer institutes, Leibniz institutes and Max Planck institutes.
Currently, TUD ranks fifth among German universities in terms of number of Horizon 2020 projects. The “Chair of Stochastic Models / Reliability, Asymptotics” belongs to the Institute of Mathematical Stochastics. This institute covers the fields stochastics, mathematical statistics, and their applications, e.g. in insurance mathematics or financial mathematics, both in teaching and in research.
The iRel4.0 partner TU Dresden’s role is contributing to work package 5: Testing 4.0 for improved Reliability. Work performed by TU Dresden is targeting the objective 5.2: AI based Outlier detection and cognitive release methods.
Hereby, the main task is development, use and calibration of neural, high dimensional and machine learning algorithms for obtaining precise and economically sensible decision rules for acceptance or rejection of wafers/product chips. TU Dresden is collaborating with Infineon Dresden.
One key contribution of the TU Dresden as a partner of the iRel4.0 project will be combining non-Gaussian statistical methods with leading edge machine learning methods for obtaining fast algorithms suitable for real time decisions and for big data. We plan to combine these statistical methods with, for example, the ML-frameworks TensorFlow and PyTorch.
Another key contribution of the TU Dresden consists in investigating the efficiency and robustness regarding the semiconductor manufacturing process. It is planned to develop a demonstrator for new, machine learning-enabled statistical methods.