Towards Weak Information Theory: Weak-Joint Typicality Decoding Using Support Vector Machines May Lead to Improved Error Exponents

TL;DR

This paper introduces a novel decoding approach using support vector machines inspired by statistical learning, which improves error exponents over traditional joint typicality decoding in discrete memoryless channels.

Contribution

It proposes a new decoding method leveraging support vector machines and a modified error definition, enhancing error exponents in channel communication.

Findings

Support vector machine-based decoding outperforms traditional methods

Modified error criteria improve decoding performance

Simulation results demonstrate better error exponents

Abstract

In this paper, the authors report a way to use concepts from statistical learning to gain an advantage in terms of error exponents while communicating over a discrete memoryless channel. The study utilizes the simulation capability of the scientific computing package MATLAB to show that the proposed decoding method performs better than the traditional method of joint typicality decoding. The advantage is secured by modifying the traditional specification of what constitutes a decoding error. This is justified by the paradigm, also used in the program of `utilizing' noisy feedback, that one ought not to declare a condition as an error if some further processing can extract useful information from it.