Achieving near-Capacity on Large Discrete Memoryless Channels

TL;DR

This paper introduces a method to enhance the capacity of large discrete memoryless channels by optimizing the input set, enabling near-capacity performance with standard binary codes without distribution shaping.

Contribution

It presents a novel algorithm based on maximizing the cut-off rate to select input subsets, improving capacity achievement in high-cardinality DMCs.

Findings

Capacity approached within tenths of a dB

Standard binary codes suffice without distribution shapers

Effective input set reduction improves performance

Abstract

We propose a method to increase the capacity achieved by uniform prior in discrete memoryless channels (DMC) with high input cardinality. It consists in appropriately reducing the input set. Different design criteria of the input subset are discussed. We develop an efficient algorithm to solve this problem based on the maximization of the cut-off rate. The method is applied to a mono-bit transceiver MIMO system, and it is shown that the capacity can be approached within tenths of a dB by employing standard binary codes while avoiding the use of distribution shapers.