[Paper Review] A Novel Index Coding Scheme and its Application to Coded Caching

Kai Wan, Daniela Tuninetti+1 more|arXiv (Cornell University)|Feb 23, 2017

Cooperative Communication and Network Coding5 citations

TL;DR

This paper proposes a novel index coding scheme based on distributed source coding and non-unique decoding that strictly outperforms composite coding. Applied to coded caching with uncoded prefetching, the scheme achieves the optimal worst-case load, matching a known outer bound and providing a new source coding-based proof of optimality for Maddah-Ali and Niesen's original scheme.

ABSTRACT

This paper proposes a novel achievable scheme for the index problem and applies it to the caching problem. Index coding and caching are noiseless broadcast channel problems where receivers have message side information.In the index coding problem the side information sets are fixed, while in the caching problem the side information sets correspond the cache contents, which are under the control of the system designer. The proposed index coding scheme, based on distributed source coding and non-unique decoding,is shown to strictly enlarge the rate region achievable by composite coding.The novel index coding scheme applied to the caching problem is then shown to match an outer bound (previously proposed by the authors and also based on known results for the index coding problem) under the assumption of uncoded cache placement/prefetching.

Motivation & Objective

• To address the suboptimality of existing inner bounds—particularly composite coding—for the index coding problem in the context of coded caching.
• To develop a new achievable scheme for index coding that leverages distributed source coding and non-unique decoding to improve rate region performance.
• To apply the new index coding scheme to the coded caching problem with uncoded cache placement and show it achieves the optimal worst-case load.
• To provide an alternative, source-coding-based proof of optimality for the Maddah-Ali and Niesen scheme under uncoded prefetching, independent of linear coding analysis.

Proposed method

• The proposed scheme uses Han's coding framework combined with Slepian-Wolf source coding and non-unique decoding to achieve improved rate regions in index coding.
• It models the index coding problem as a source coding problem with side information, where messages are encoded using structured linear combinations based on user demands and side information.
• The scheme defines auxiliary random variables and transmission rates via entropy-based bounds, with key equations involving mutual information and conditional entropy to characterize achievable rates.
• The transmission strategy involves sending linear combinations of file subfiles, where each packet corresponds to a subset of users' demands and side information.
• The symmetric rate is derived as $ R_{\text{sym}} = \frac{1}{\binom{K}{t+1} - \binom{K - \min(N,K)}{t+1}} \log_2(|\mathcal{X}|) $, ensuring rate optimality.
• The method is validated by showing that the symmetric rate matches the known outer bound under uncoded cache placement, proving optimality for the worst-case load.

Experimental results

Research questions

• RQ1 Can a new index coding scheme based on distributed source coding and non-unique decoding outperform composite coding in terms of achievable rate region?
• RQ2 Does the proposed index coding scheme achieve the optimal worst-case load in coded caching systems with uncoded prefetching?
• RQ3 Can the optimality of the Maddah-Ali and Niesen scheme be re-proven using a source coding-with-side-information framework rather than linear coding analysis?
• RQ4 Is the proposed scheme universally applicable to general index coding problems beyond caching?
• RQ5 What is the precise characterization of the symmetric rate achieved by the new scheme in terms of system parameters such as number of users, files, and cache size?

Key findings

• The proposed index coding scheme strictly improves upon composite coding, as demonstrated by a concrete example where the new scheme achieves a lower rate region.
• The scheme achieves the symmetric rate $ R_{\text{sym}} = \frac{1}{\binom{K}{t+1} - \binom{K - \min(N,K)}{t+1}} \log_2(|\mathcal{X}|) $, which matches the known outer bound for the worst-case load.
• For the caching problem with uncoded placement, the scheme achieves the optimal worst-case load, confirming the optimality of the Maddah-Ali and Niesen scheme under this constraint.
• The symmetric rate is shown to be achievable via mutual information bounds, with $ R_{\text{sym}} = \frac{1}{3} \log_2(|\mathcal{X}|) $ in the example with 6 users and 3-bit transmissions.
• The sum-rate delivered to each user is $ R_{\text{sum-rate}} = \frac{\binom{K}{t}}{\binom{K}{t+1} - \binom{K - |\mathcal{N}(\mathbf{d})|}{t+1}} \log_2(|\mathcal{X}|) $, and the inverse of this rate corresponds exactly to the load in the original caching problem.
• The scheme provides a new source-coding-based proof of optimality for the original coded caching scheme, independent of linear code analysis, and applies to both centralized and decentralized caching systems.