Maximilian Stark received the Karl H. Ditze award for his master thesis “Information Optimum Design of Discrete LDPC Decoders for Irregular Codes” which he had done at the Institute of Communications (INT) under the supervision of Jan Lewandowsky. The award ceremony with participation of Prof. Dr.-Ing. Rolf Dalheimer, board member of the Karl H. Ditze Foundation, took place on July 11, 2018 at TUHH.

The award winners: Richard Bäumer, Anne Rynek, Maximilian Stark

The master thesis considers an innovative approach to “information-optimum” signal processing using the example of an error correcting LDPC code. In digital wireless communications systems, the error correcting decoder is the by far computationally most expensive baseband signal processing component. Therefore, an energy- and computationally efficient solution is a major challenge in reliable high data rate and low latency transmission as required e.g. in 5G mobile communications, autonomous driving or industry 4.0 applications.

State of the art methods are usually derived from an optimum algorithm which is made implementable by approximations and reduced bit width. Approximations are “local” in the sense that in each stage of the signal processing chain, the signal samples and operations are quantized and approximated as good as possible. In contrast, Maximilian Stark pursues the concept of preserving “relevant information”: Quantization and operations are designed using the so-called information bottleneck method such that a maximum of relevant mutual information in the sense of Shannon’s information theory is preserved. 

It is interesting that the information bottleneck method uses information-theoretic measures but is hardly known in communications engineering or signal processing. It has been mainly used in the machine learning community for problems such as text classification or in neurobiological applications. The translation of the information bottleneck method to communications engineering problems can, therefore, be considered to be an interdisciplinary achievement.

In prior work at the institute of communications, the concept had been applied to decoding of regular LDPC codes. However, irregular codes are more widely used in practice due to their superior error correction capabilities.

The contribution of Maximilian Stark is the development of the message alignment principle for information-optimum decoding of irregular LDPC codes which avoids the problem of ambiguities in the bit reliabilities.

Using computer simulations, he shows the remarkable result that compared to state-of-the-art decoders, superior error rate performance can be achieved at lower hardware complexity.

The Karl H. Ditze foundation was founded in 1979 for promotion of the universities in Hamburg. At Hamburg University of Technology (TUHH), the Ditze foundation sponsors three  yearly awards: One for the best bachelor thesis, master thesis and Ph.D. thesis, respectively. Besides scientific excellence, practical significance and multidisciplinarity are important evaluation criteria.