Motivation

In an Information Bottleneck (IB) decoder all internal functions are designed via IB to process quantized signals with a very low bit resolution (e.g. 3-4 bit). In a straight forward implementation, these functions are usually implemented as Lookup-Tables (LUT‘s). However, the memory requirement of design and storing is exponential in the number of inputs. To overcome this issue, a common approach is to utilize a simple LUT decomposition that usually comes with the cost of BER/FER performance loss.

Objective

Investigation of discrete functions and/or LUT designs that lead to efficient discrete decoder implementations.

Requirements

Channel coding, basic programming skills