ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let (U,V)→X→(Y,Z)(U,V)\to X\to(Y,Z) be a Markov chain for a broadcast channel with binary input alphabet ∣X∣=2|\mathcal{X}|=2. Define

Tα(X):=max⁡p(u,v∣x)αI(U;Y)+I(V;Z)−I(U;V),α≥1,T_\alpha(X):=\max_{p(u,v\mid x)}\alpha I(U;Y)+I(V;Z)-I(U;V),\qquad \alpha\geq1,

and let C[f(X)]\mathfrak{C}[f(X)] denote the upper concave envelope of ff at p(x)p(x). Put λˉ=1−λ\bar\lambda=1-\lambda.

Binary-alphabet concave-envelope conjecture. For all α≥1\alpha\geq1 and all such Markov chains,

−(α−λˉ)H(Y)−λˉH(Z)+Tα(X)≤C[−(α−λˉ)H(Y)−λˉH(Z)+max⁡{αI(X;Y),I(X;Z)}].-(\alpha-\bar\lambda)H(Y)-\bar\lambda H(Z)+T_\alpha(X) \leq \mathfrak{C}[-(\alpha-\bar\lambda)H(Y)-\bar\lambda H(Z)+\max\{\alpha I(X;Y),I(X;Z)\}].

The underlying pointwise inequality is known to fail for some binary-input broadcast channels, but the conjectured concave-envelope inequality may still hold for the distributions needed to compute the envelope. The source proves the underlying inequality for several cases and for the binary skew-symmetric broadcast channel, but gives no resolution of this conjecture.

Record

Comments

Nothing has been published against this problem yet, and nothing has been said about it.

Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.

Sign in

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.