Table of Contents
Fetching ...

Indirect Lossy Source Coding with Observed Source Reconstruction: Nonasymptotic Bounds and Second-Order Asymptotics

Huiyuan Yang, Yuxuan Shi, Shuo Shao, Xiaojun Yuan

TL;DR

This paper considers the joint compression of a pair of correlated sources, where the encoder is allowed to access only one of the sources, and derives nonasymptotic achievability and converse bounds valid for general sources and distortion measures.

Abstract

This paper considers the joint compression of a pair of correlated sources, where the encoder is allowed to access only one of the sources. The objective is to recover both sources under separate distortion constraints for each source while minimizing the rate. This problem generalizes the indirect lossy source coding problem by also requiring the recovery of the observed source. In this paper, we aim to study the nonasymptotic and second-order asymptotic properties of this problem. Specifically, we begin by deriving nonasymptotic achievability and converse bounds valid for general sources and distortion measures. The source dispersion (Gaussian approximation) is then determined through asymptotic analysis of the nonasymptotic bounds. We further examine the case of erased fair coin flips (EFCF) and provide its specific nonasymptotic achievability and converse bounds. Numerical results under the EFCF case demonstrate that our second-order asymptotic approximation closely approximates the optimum rate at appropriately large blocklengths.

Indirect Lossy Source Coding with Observed Source Reconstruction: Nonasymptotic Bounds and Second-Order Asymptotics

TL;DR

This paper considers the joint compression of a pair of correlated sources, where the encoder is allowed to access only one of the sources, and derives nonasymptotic achievability and converse bounds valid for general sources and distortion measures.

Abstract

This paper considers the joint compression of a pair of correlated sources, where the encoder is allowed to access only one of the sources. The objective is to recover both sources under separate distortion constraints for each source while minimizing the rate. This problem generalizes the indirect lossy source coding problem by also requiring the recovery of the observed source. In this paper, we aim to study the nonasymptotic and second-order asymptotic properties of this problem. Specifically, we begin by deriving nonasymptotic achievability and converse bounds valid for general sources and distortion measures. The source dispersion (Gaussian approximation) is then determined through asymptotic analysis of the nonasymptotic bounds. We further examine the case of erased fair coin flips (EFCF) and provide its specific nonasymptotic achievability and converse bounds. Numerical results under the EFCF case demonstrate that our second-order asymptotic approximation closely approximates the optimum rate at appropriately large blocklengths.
Paper Structure (21 sections, 17 theorems, 171 equations, 5 figures)

This paper contains 21 sections, 17 theorems, 171 equations, 5 figures.

Key Result

Proposition 1

$\tilde{\mathcal{V}}(d_s,d_x)$ can be written as where $\textrm{Var}\left[U|V\right] \triangleq \mathbb{E}\left[(U - \mathbb{E}\left[U|V\right])^2\right]$.

Figures (5)

  • Figure 1: Indirect lossy source coding with observed source reconstruction in the nonasymptotic regime.
  • Figure 2: The five regions of the distortion plane with $\delta = 0.2$.
  • Figure 3: The rate-distortion function of the EFCF case with $\delta = 0.2$.
  • Figure 4: The rate dispersion function under the EFCF case with $\delta = 0.2$.
  • Figure 5: Rate-blocklength trade-off in the erased fair coin flips case with $\delta = 0.2$ and $\epsilon = 0.1$. Note that $(d_x, d_s) = (1.36\delta, 0.88\delta)$ and $(d_x, d_s) = (1.28\delta, 1.12\delta)$ share the same value of $R(d_s,d_x)$.

Theorems & Definitions (40)

  • Definition 1
  • Definition 2
  • proof : Proof
  • Remark 1
  • Proposition 1
  • proof : Proof
  • Theorem 1
  • proof : Proof
  • Theorem 2
  • proof : Proof
  • ...and 30 more