Moderate Deviation Analysis for Classical-Quantum Channels and Quantum Hypothesis Testing

Hao–Chung Cheng, Min-Hsiu Hsieh

Abstract

In this work, we study the tradeoffs between the error probabilities of classical-quantum channels and the blocklength $n$ when the transmission rates approach the channel capacity at a rate slower than $1/\sqrt{n}$, a research topic known as moderate deviation analysis. We show that the optimal error probability vanishes under this rate convergence. Our main technical contributions are a tight quantum sphere-packing bound, obtained via Chaganty and Sethuraman's concentration inequality in strong large deviation theory, and asymptotic expansions of error-exponent functions. Moderate deviation analysis for quantum hypothesis testing is also established. The converse directly follows from our channel coding result, while the achievability relies on a martingale inequality.

Introduction: The Three Regimes of Information Theory

In information theory, communication performance is typically analyzed through the lens of error probability tradeoffs across three distinct regimes. These regimes are defined by how the transmission rate approaches the channel capacity $C$ as the blocklength $n$ increases:

  1. Small Error (Large Deviation): The transmission rate is fixed below capacity. Here, the error probability vanishes exponentially as $n \to \infty$.
  2. Large Error (Second-Order): The rate approaches capacity very quickly (at a scale of $1/\sqrt{n}$), leading to a constant, non-vanishing error probability governed by the Central Limit Theorem (CLT).
  3. Medium Error (Moderate Deviation): The rate approaches capacity at a "moderate" speed—slower than $1/\sqrt{n}$ but faster than a constant.
Comparison of error regimes across concentration phenomena, hypothesis testing, and channel coding.

While the first two regimes were well-studied for classical-quantum (c-q) channels—where inputs are classical but outputs are quantum states—the moderate deviation regime remained largely unexplored. This paper bridges that gap, providing a precise characterization of the error decay in this intermediate zone.

Methodology: Bridging the Gap with Strong Large Deviation Theory

The primary challenge in moderate deviation analysis is finding a mathematical bound tight enough to ensure that achievability (what is possible) and the converse (what is impossible) meet at the same value. The authors employ several advanced tools:

  • Achievability: Starting with Hayashi’s upper bound for average error in c-q channels, the authors apply a Taylor expansion to the error-exponent function around the channel capacity to derive the moderate deviation behavior.
  • Converse: The paper's main technical contribution is a sharp quantum sphere-packing bound. By utilizing the Chaganty-Sethuraman concentration inequality from strong large deviation theory, the authors achieve a level of precision that standard quantum bounds cannot provide.
  • Martingale Analysis: For quantum hypothesis testing, the authors provide an alternative achievability proof using a refined Azuma's inequality, treating the quantum state discrimination task as a sequence of random variables.

Key Results: The Moderate Deviation Property

The central result of this work is the characterization of the optimal error probability $\epsilon^*$ for a c-q channel with capacity $C$ and dispersion $V$. If the rate approaches capacity as $C - a_n$ (where $a_n$ is a sequence vanishing slower than $1/\sqrt{n}$), the error probability follows:

$$\lim_{n\to\infty} \frac{\log \epsilon^*(n, C - a_n)}{n a_n^2} = -\frac{1}{2V}$$

This formula demonstrates that the error probability decays as $\exp(-n a_n^2 / 2V)$. It reveals that the "penalty" for approaching capacity in this regime is governed entirely by the channel's dispersion $V$, a quantity usually associated with the second-order (CLT) regime.

Furthermore, the authors show a symmetry in hypothesis testing: the error in distinguishing two quantum states $\rho$ and $\sigma$ follows the same mathematical structure, where the decay rate is determined by the relative entropy variance.

Conclusion and Future Directions

This research successfully closes the gap in the moderate deviation analysis for c-q channels by showing that the Moderate Deviation Property (MDP) acts as a bridge between the CLT and Large Deviation Principle (LDP) regimes.

Future work remains in extending these tight results to fully quantum channels (quantum input and output), where entanglement adds significant complexity. Additionally, while this work relies on LDP-based techniques, reconciling these findings with contemporary CLT-based methods could lead to a more unified framework for quantum asymptotic information theory.

Summary updated

Abstract

In this paper, we study the tradeoffs between the error probabilities of classical-quantum channels and the block-length n when the transmission rates approach the channel capacity at a rate lower than 1/√n, a research topic known as moderate deviation analysis. We show that the optimal error probability vanishes under this rate convergence. Our main technical contributions are a tight quantum sphere-packing bound, obtained via Chaganty and Sethuraman's concentration inequality in strong large deviation theory, and asymptotic expansions of error-exponent functions. Moderate deviation analysis for quantum hypothesis testing is also established. The converse directly follows from our channel coding result, while the achievability relies on a martingale inequality.
BibTeX
@article{cheng2017moderate,
  author = {Hao–Chung Cheng and Min-Hsiu Hsieh},
  title = {{Moderate Deviation Analysis for Classical-Quantum Channels and Quantum Hypothesis Testing}},
  journal = {IEEE Transactions on Information Theory},
  year = {2017},
  doi = {10.1109/tit.2017.2781254},
}
Figure