Tradeoff between decoding complexity and rate for codes on graphs

TL;DR

This paper investigates the tradeoff between decoding complexity and achievable rate for graph-based codes like LDPC and LDGM, establishing bounds and structural conditions for near-capacity performance with bounded complexity.

Contribution

It provides a lower bound on the achievable rate under bounded decoding complexity and identifies structural constraints of codes that approach capacity.

Findings

Achievable rate is bounded below capacity with finite decoding steps.

Good low-complexity performance implies strong local graph structures.

Decoding complexity scales at least as O(log(1/epsilon)) near capacity.

Abstract

We consider transmission over a general memoryless channel, with bounded decoding complexity per bit under message passing decoding. We show that the achievable rate is bounded below capacity if there is a finite success in the decoding in a specified number of operations per bit at the decoder for some codes on graphs. These codes include LDPC and LDGM codes. Good performance with low decoding complexity suggests strong local structures in the graphs of these codes, which are detrimental to the code rate asymptotically. The proof method leads to an interesting necessary condition on the code structures which could achieve capacity with bounded decoding complexity. We also show that if a code sequence achieves a rate epsilon close to the channel capacity, the decoding complexity scales at least as O(log(1/epsilon).