Jan Lewandowsky received the Sick award for his phd thesis „The Information Bottleneck method in Communications.” The online award ceremony with participation of the donator Renate Sick-Glaser as well as the SICK foundation board member Wolfgang Bay took place on July 13, 2021.
Energy-efficient signal processing is more and more becoming a major challenge in high data rate applications such as wireless or fiber optical communications systems or the internet of things. From the signal processing point of view, reduced complexity algorithmic implementations and coarse quantization, i.e. a signal representation with a very small number of bits per sample, are key approaches.
Unfortunately, reduced complexity implementations and coarse quantization typically result in significant performance degradations with respect to optimum solutions.
This drawback is addressed in the thesis by Jan Lewandowsky. Instead of targeting a best possible signal value approximation, he pursues the concept of preserving an adequately defined relevant information in the sense of Shannon’s information theory. He applies the so-called Information Bottleneck Method, a classification tool from the machine learning community, in order to design coarse quantizers which preserve relevant information throughout the whole signal processing chain. This leads to fundamentally different quantizers compared to the state of the art. Moreover, complex operations degenerate to simple look-up tables. The remarkable result is that excellent performance extremely close to the theoretical case of unquantized signal processing can be achieved in many relevant cases such as forward error control decoding and channel estimation.
The thesis covers the broad range from information-theoretical fundamentals, the enhancement of algorithms, the development, optimization and numerical evaluation of an innovative information processing concept up to a DSP implementation.
It is also excellent from a didactical point of view and contains many examples which help the reader to catch the abstract content.
The annual SICK awards (SICK Wissenschaftspreise) are donated by the Gisela und Erwin Sick Foundation. There is one award each for the best bachelor thesis, the best master thesis and the best dissertation at TUHH. The Gisela und Erwin Sick Foundation was established in 2008 by Gisela Sick. She was the wife of Erwin Sick who founded the SICK AG in 1946. As a sensor manufacturer, the company is a technology and market leader providing sensors and application solutions for factory automation, logistics automation and process automation.