Information Bottleneck Message Passing for Military Applications

Abstract

Military communication systems naturally have strong requirements concerning the reliability and the robustness of their physical layer data transmission schemes. Modern channel coding and modulation schemes can meet these requirements in general, but their detection and decoding at the receiving end requires complex and power demanding high-precision implementations of digital algorithms, which are often not suitable for military applications. This motivates to explore novel techniques to build simple detection and decoding algorithms with good performance. In this article, we present novel results on the recent idea of using a machine learning framework termed the Information Bottleneck method to replace demanding implementations of the sum-product algorithm with very simple quantized message passing schemes. We provide a novel explanation, which links the Information Bottleneck decoder processing to the sum-product algorithm. Moreover, we present a novel Information Bottleneck demodulation scheme for quadrature amplitude modulation and discuss special advantages of the Information Bottleneck system design approach for military applications.