PaperlayerView exact revision on arXiv

Source-linked AI summary

Universality Sacrifices Reliability in Classical-Quantum Channel Coding

Kaito Watanabe, Masahito Hayashi, Takaya Matsuura, Hao-Chung Cheng

arXiv:2610.01941v1quant-phcs.ITmath.IT

TL;DR

Universal coding must operate without a complete channel description, but whether it preserves optimal reliability for classical-quantum channels is unresolved by classical results. The paper proves a reliability loss can occur under unknown output rotations and constructs a channel-independent scheme whose bound matches the unitary-invariant converse at high rates.

  • Problem

    Whether universal classical-quantum coding can attain channel-aware optimal reliability despite lacking the channel description, including the orientation of its output states, is the central question.

    Source: p. 1

  • Method

    The paper exhibits a phase-rotated channel family, derives a reliability converse for unitary-invariant decoders, and constructs a channel-independent encoder-decoder pair with a universal random-coding bound.

    Source: pp. 1–3

  • Results

    For noncommuting outputs the channel-aware and universal reliability exponents can differ, governed respectively by Petz and sandwiched Rényi divergences; the universal bound matches the unitary-invariant converse at high rates.

    Source: p. 2

  • Takeaways & Limitations

    The output system's unknown unitary orientation can limit the achievable error exponent, whereas the two divergences coincide for commuting outputs.

    Source: p. 2

    For the phase-rotated family, the phase witnessing nonattainment may depend on the code sequence and rate, but not on blocklength.

    Source: p. 3, End Matters · p. 8

Abstract

from arXiv · show

Universal channel coding enables communication without a complete description of the channel. For classical channels, universal codes can attain both capacity and the optimal high-rate reliability. We show that this compatibility fails for classical-quantum channels in general; that is, the optimal reliability in the channel-aware scenario is not always achievable with universal coding due to the ignorance of the unitary rotation of the output system. We exhibit a family of classical-quantum channels for which one cannot achieve the channel-aware optimal reliability by a fixed coding scheme. We further derive a converse bound on the reliability for unitary-invariant decoders, a natural assumption for the universal coding scheme, that can be strictly smaller than the optimal channel-aware error exponent. Conversely, we construct a channel-independent encoder-decoder pair and establish a universally achievable bound on the reliability that matches this converse bound in the high-rate regime, thereby characterizing the optimal universal reliability. Specifically, the channel-aware and universal exponents are governed by the Petz and sandwiched R\'enyi divergences, respectively. These divergences coincide for commuting outputs but differ for noncommuting ones, explaining why universality preserves optimal reliability classically but can reduce it quantumly. Our results showcase the fundamental reliability cost of performing the classical-quantum channel coding task universally.

End Matters

A phase-rotated qubit-channel family shows that channel-aware random-coding reliability cannot in general be attained by a fixed channel-independent coding scheme. The no-go result compares the universal-code exponent with the channel-aware exponent and identifies a phase that may depend on the code sequence and rate, but not on blocklength.

End Matters

  • A concrete phase-rotated channel family demonstrates that the channel-aware Petz random-coding bound is not universally achievable.The outputs are noncommuting, and the same unknown phase is used at every channel use.Source: End Matters · p. 7
  • For the exhibited channel family, every rate-R code sequence independent of the phase has a reliability bound strictly below the channel-aware random-coding exponent.The strict comparison is D(t||p) < E_aware(R), with h(t)=R.Source: End Matters · p. 7
  • The comparison concerns positive rates below h(p), with t in (0,p) chosen so that h(t)=R.The phase is constant throughout each block.Source: End Matters · p. 7
  • The fixed-phase result rules out attainment for every fixed phase, while the phase witnessing failure may vary with the code sequence and rate, but not with blocklength.The proof uses the strict gap established for the phase-rotated family.Source: End Matters · p. 8

Appendices

The appendices set out the information quantities and combinatorial tools used in the paper. They define classical and quantum divergences and mutual information, then review types and conditional types for finite sequences.

1. Entropic quantities

  • The appendix defines classical and quantum relative entropies, including the Petz and sandwiched Rényi divergences, and notes that both quantum divergences converge to Umegaki relative entropy as α approaches 1.It also introduces measured and quasi-measured Rényi quantities.Source: §1 · pp. 9–10
  • Classical mutual information is defined from a joint distribution, and the appendix introduces its classical-quantum counterpart for an input distribution and channel.The classical-quantum definition follows the discussion of quantum relative entropies.Source: §1 · p. 10

2. Type theory

  • A type is the empirical distribution of a sequence, and a type class contains the sequences with that type.The appendix reviews the number of types and bounds on type-class cardinalities.Source: §2 · pp. 10–11
  • A conditional type compatible with an input type specifies conditional output frequencies; values outside the input type's support are immaterial.Conditional types agreeing on the support are identified.Source: §2 · p. 11

Appendix B: Sphere-packing bound under unitary-invariance (Proof of Theorem 1)

This appendix derives the sphere-packing bound for decoders invariant under collective unitary transformations. The stated result applies to constant-composition codebooks and identifies a high-rate regime for the bound.

Appendix B: Sphere-packing bound under unitary-invariance (Proof of Theorem 1)

  • The appendix derives a sphere-packing bound under the decoder constraint [Λ_m, U^⊗n]=0 for every unitary U and message m.The goal is a converse bound for unitary-invariant decoding.Source: Appendix B · p. 11
  • Theorem S.1 gives the bound for constant-composition codebooks whose n-types converge to a fixed input distribution P.It is stated for any coding scheme with a unitary-invariant decoder.Source: Appendix B · p. 11
  • The bound is used in the high-rate regime R at or above a critical rate, and the associated quantity is the sandwiched Augustin information.The excerpt specifies R_crit ≤ R.Source: Appendix B · p. 11

1. Reduction to the pinched hypothesis testing

The proof reduces a coding converse to quantum hypothesis testing, then uses unitary invariance to relate testing on pinched states to unrestricted testing. A divergence limit supplies the sandwiched-Rényi quantity used in the subsequent argument.

1. Reduction to the pinched hypothesis testing

  • In quantum hypothesis testing, a binary measurement distinguishes ρ from σ, with type I and type II errors corresponding to the two possible wrong decisions.The type I probability is minimized subject to a bound on the type II probability in the operational quantity described in the passage.Source: §1 · p. 11
  • For unitary-invariant tests, pinching preserves the type I error, and the constrained minimum is no smaller than the minimum over all tests.The latter minimization ranges over a superset of the invariant tests.Source: §1 · p. 12
  • A proposition connects decoding error for a classical-quantum code with hypothesis-testing error when the decoder is unitary-invariant.The proposition allows any channel and fixed comparison state σ.Source: §1 · p. 12
  • The appendix relates the regularized divergence of a pinched state to the sandwiched divergence of the original state for a sequence of types converging to P.The comparison state is assumed full rank.Source: §1 · pp. 12–13

2. Converse bound from large deviation

The section develops a converse for unitary-invariant decoders using large-deviation results for pinched states. It also establishes convergence and regularity properties of the associated log-moment functions.

2. Converse bound from large deviation

  • The limiting log-moment function is a convex combination of single-letter sandwiched log-moment functions, with pinching controlling finite-n differences.The component functions are finite, convex, and differentiable on (0, 1), supplying log-moment conditions for the Gärtner–Ellis argument.Source: §2 · pp. 13–14
  • The pinched-state Hoeffding exponent is the large-deviation result used in the converse argument.The passage identifies this as Proposition S.5 for a full-rank state and types converging to P.Source: §2 · p. 14
  • The fixed-auxiliary-state converse applies to every sequence of constant-composition codes with unitary-invariant decoders.It assumes a positive target rate, a convergent sequence of types, and a full-rank auxiliary state.Source: §2 · p. 14

3. Sandwiched sphere-packing bound via minimax

This section establishes a minimax identity for the sandwiched sphere-packing optimization and connects its optimized form to the sandwiched Augustin information. The rate range for the latter connection is restricted by the minimax argument.

3. Sandwiched sphere-packing bound via minimax

  • The minimax step justifies exchanging the infimum and supremum and extending the state optimization range needed for the sandwiched sphere-packing bound.The objective is convex and lower semicontinuous in the state for 0 < s ≤ 1, and concave in s for fixed state.Source: §3 · pp. 14–15
  • Proposition S.7 asserts a minimax identity for any classical-quantum state W mapping x to ρx; the cited passage does not specify the identity's formula.The infimum and supremum are attained, and their optimizers form a saddle point.Source: §3 · p. 15
  • The saddle-point argument establishes attainment of the supremum by showing the objective is upper semicontinuous on [0, 1].The section notes that the limit at s approaching zero equals the value assigned at zero.Source: §3 · pp. 14–15

The right-hand side: Sandwiched Augustin information

  • The sandwiched Augustin-information formulation applies over the rate region Rcrit(P) < R < I(P;W), where maximizers lie inside (0, 1).The endpoint derivative signs exclude both endpoints for rates in this interval.Source: The right-hand side: Sandwiched… · p. 15

The left-hand side: Hoeffding divergence

This section strengthens the minimax result and uses perturbation of a possibly singular optimizer to support the high-rate sphere-packing bound. The perturbation makes the auxiliary state full-rank before a limit is taken.

The left-hand side: Hoeffding divergence

  • The strengthened result follows by establishing a saddle point and a unique interior maximizer for the fixed-state objective.The proof uses concavity and real analyticity to establish uniqueness.Source: The left-hand side: Hoeffding divergence · p. 16
  • A potentially singular optimizer is perturbed to a full-rank state, and the perturbation is then taken to zero to obtain the high-rate bound.For Rcrit(P) < R < I(P;W), the proof combines the perturbation lemma with the minimax proposition.Source: The left-hand side: Hoeffding divergence · pp. 16–17
  • The proof confines global maximizers to a compact interval using derivative inequalities that persist under small perturbations and nearby rates.Continuity of the objective and its first derivative on the chosen interval ensures the maximum is attained there.Source: The left-hand side: Hoeffding divergence · pp. 16–17
  • The optimizer’s support contains the states ρx for inputs with P(x) > 0, allowing analysis on the support before perturbation.On that support, the restricted optimizer is positive definite.Source: The left-hand side: Hoeffding divergence · p. 16

Appendix C: Proof of Theorem 2

The appendix proves a universal random-coding bound by constructing a fixed constant-composition codebook and a channel-independent decoder. Its proof connects decoder error bounds to measured and then sandwiched Augustin information.

Appendix C: Proof of Theorem 2

  • A fixed sequence of constant-composition codebooks and decoders satisfies a decoding-error bound for every classical-quantum channel.The theorem states the bound for codebook types converging to P and rates in the specified range.Source: Appendix C · p. 18

2. Construction of the decoder

  • The decoder proof first bounds error for a fixed good codebook, then uses a channel-independent decoder to relate error to measured Augustin information.The regularized measured Augustin information is then identified with the sandwiched Augustin information.Source: Appendix C · p. 18

1. Construction of the codebook

  • The construction selects good codebooks from a sequence whose existence is guaranteed for types converging to P.The appendix takes that sequence as its codebook.Source: §1 · pp. 18–19

Quotient decoder

  • The quotient-decoder construction forms a complete POVM from codeword-associated operators, with a covariance condition enabling an error upper bound.The proposition applies to a good codebook and a family of positive definite operators satisfying that condition.Source: Quotient decoder · p. 19

Measured R´enyi Augustin information and its duality

This section develops variational and dual descriptions of measured Rényi Augustin information and uses positive operators to bound decoding error.

Measured R´enyi Augustin information and its duality

The universal family of positive operators

This section constructs a family of positive operators independent of the channel description, enabling a universal decoder and a decoding-error bound.

The universal family of positive operators

Regularization from the measured to the sandwiched R´enyi Augustin information

This section regularizes the measured Rényi Augustin information to the sandwiched version, supporting the universal bound used in the main theorem.

Regularization from the measured to the sandwiched R´enyi Augustin information

Combining everything altogether: Proof of Theorem S.9 through double-blocking

The proof of Theorem S.9 uses double-blocking to apply the universal decoder construction and obtain the claimed reliability bound.

Combining everything altogether: Proof of Theorem S.9 through double-blocking

Loading 2610.01941v1…