Motivation:

The decoding of short block length codes is a challenging task and the question naturally arises whether a purely data driven decoder based on neural networks (NN) might be an alternative. There already has been some research ongoing in this direction and first results show that these NN-based decoder can yield good performance for short codes but fail if the number of possible codewords grow too large. This shall be further investigated in this project thesis.

Goal:

The aim of this thesis is to implement an NN-based decoder for very short block length and further investigate the mentioned results. Therefore investigations about machine learning libraries and literature research needs to be done.

Requirements:

In order to process this thesis, knowledge of Channel Coding 1 and 2, Wireless Communication Technologies and programming skills in Python are essential.