Soft Covering Through the Lens of Hypothesis Testing

2026-05-19

AI Summary

This study addresses the hypothesis testing problem of distinguishing whether a channel output sequence is generated by a random codeword or independently drawn from its marginal distribution. Within the Neyman–Pearson framework, the authors leverage large deviation theory, random coding arguments, and single-letter information-theoretic techniques to derive tight single-letter expressions for the error exponents governing the exponential decay rates of false alarm and missed detection probabilities. The central contribution lies in uncovering a phase-transition structure inherent in soft covering: as the coding rate crosses the mutual information threshold, the joint behavior of the two error exponents undergoes a sharp qualitative change, thereby fully characterizing the phase-transition phenomenon in error exponents as a function of code rate.

Abstract

We study the soft covering phenomenon through the lens of Neyman--Pearson hypothesis testing: given a channel output sequence $y^n$, can one decide whether it was produced when the channel was driven by a random codeword, or generated independently from the output marginal? We derive exact exponential decay rates for the jointly averaged false-alarm (FA) probability $α_n(τ,R)$ and missed-detection (MD) probability $β_n(τ,R)$, as functions of the decision threshold $τ$ and the codebook rate $R$. The derived single-letter formulas of the exponents $\EFA(τ,R)=-\lim_{n\to\infty}\frac{1}{n}\lnα_n(τ,R)$ and $\EMD(τ,R)=-\lim_{n\to\infty}\frac{1}{n}\lnβ_n(τ,R)$ are tight in the random coding sense. The analysis reveals a rich phase structure. For $R < I(X;Y)$, there is a genuine exponential tradeoff between the two error types over the interval $τ\in (0, I(X;Y)-R)$. At $R = I(X;Y)$, this tradeoff interval collapses to the single point $τ= 0$, where both error exponents simultaneously vanish, a fact which manifests the soft covering phenomenon in the Neyman--Pearson sense. For $R > I(X;Y)$, the same instantaneous collapse persists at $τ= 0$; moreover, for every $τ$ at least one exponent is zero: the FA exponent is zero for $τ\le 0$ (FA probability does not decay exponentially), and the MD exponent is zero for $τ\ge 0$ (and finite, channel-specific for $τ<0$; see Remark~\ref{rem:jump}). There is no interval of $τ$ where both exponents are simultaneously positive. A sharp phase transition in the MD exponent occurs at $τ^* = [I(X;Y)-R]_+$ for all rates.

Problem

• soft covering
• hypothesis testing
• error exponents
• Neyman-Pearson
• channel coding

Innovation

• soft covering
• hypothesis testing
• error exponents
• phase transition
• random coding