arXiv:2610.41858
Channel dichotomies
Aadil Oufkir, Mario Berta
Abstract
We study channel interconversion under non-signalling superchannels, where the asymptotic conversion rate is given by the ratio of the channel capacities. Our main result determines the second-order expansion of the optimal target length. Namely, when each channel has a single capacity-achieving dispersion, the coefficient is given by the inverse sesquinormal distribution, with parameters determined by the capacities and dispersions. The result extends state-dichotomy theory to channels: one superchannel approximates the target uniformly over inputs and maps every reference channel whose output is independent of its input to another such channel. As in state dichotomies, there are finite-error resonances where the second-order term vanishes, leaving an $O(\log n)$ remainder. Crucially, our new characterization cannot be obtained by concatenating channel coding and simulation protocols [Cao {\it et al.}~IEEE Trans.~Inf.~Theory (2024)]. Instead, the proof combines comparisons of binary hypothesis testing experiments with uniform Gaussian estimates and a construction that uses symmetry within type classes to satisfy the non-signalling constraints. Shared randomness assistance already suffices at first order, but we show that its second-order behaviour can differ strictly from the non-signalling case.
AI Summary
The paper studies how many uses of a finite classical target channel $W$ can be simulated from $n$ uses of a resource channel $T$, under non-signalling (NS) assistance and worst-input total-variation error $\epsilon$. It determines the fixed-error second-order correction to the capacity-ratio rate, rather than separately optimizing channel coding and simulation. When each channel has a single positive capacity-achieving dispersion, the optimal target length is $m^*_{\mathrm{NS}}=rn+(d_\epsilon/C_W)\sqrt n+O(\log n)$, where $r=C_T/C_W$ and $d_\epsilon=\inf_{\epsilon<\eta<1}\{\sqrt{V_T}\Phi^{-1}(\eta)-\sqrt{rV_W}\Phi^{-1}(\eta-\epsilon)\}$. This is the inverse sesquinormal profile known from state-dichotomy conversion. The limiting error at the capacity-ratio length vanishes when $V_T/C_T=V_W/C_W$; otherwise, a positive threshold error cancels the second-order term. A generalization treats nonunique positive dispersions through an optimization over resource Gaussian mixtures and target variances. The converse uses binary hypothesis-testing comparisons with uniform bounds covering correlated inputs and protocol-dependent reference distributions. Achievability combines type-class symmetrization with explicit completion of both NS marginals, paying $O(\log n)$ resource uses for auxiliary communication. Exact finite-blocklength formulas are also derived for additive channels and BEC-to-BSC conversion. Shared randomness attains the NS second-order profile for packet-erasure-to-bit-erasure conversion and BSC self-conversion, but a dispersion-matched BEC-to-BSC example exhibits a strict separation. General shared-randomness and entanglement-assisted characterizations remain open, as do general nonadditive zero-dispersion cases and third-order refinements.
Based on previous tight second-order results for the interconversion of specific channels, the general characterization given here was developed with ChatGPT (5.6 Sol and 6 Astra).
Error found: The abstract states the singleton-dispersion characterization without the strictly positive capacities and dispersions required by the main theorem, while the paper does not establish the claimed generality for nonadditive zero-dispersion channels.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 34 pages, 4 figures