Research Interests

My current research involves the development of innovative tensor methods and their application on high-dimensional data problems in signal processing, machine learning, system identification and control. For example, I have developed and implemented an algorithm that uses tensor networks to estimate the coefficients of a multiple-input-multiple-output (MIMO) Volterra model from measured data. This algorithm is able to estimate 10 million coefficients in about 1 second on a standard desktop computer, whereas the identified model is able to predict future data with a relative error of 5%. Other developed tensor algorithms are a Tensor Network Kalman filter, a polynomial classifier for the classification of images, and various tensor decomposition algorithms with applications in image processing and nonlinear block-structured system identification. All my implementations are released as open source and can therefore be freely downloaded, used and modified. Furthermore, I also have a strong passion for education and the popularization of science and engineering.