ProbXiv
Explorersign in
Problem archiveProblem record

Statement

By Zamir-Feder, for any matrix AA with

X=[X1X2…Xn], and X~=[X~1X~2…X~n],X = \begin{bmatrix} X_1 & X_2 & \ldots & X_n \end{bmatrix}, \quad \text{ and } \quad \tilde{X} = \begin{bmatrix} \tilde{X}_1 & \tilde{X}_2 & \ldots & \tilde{X}_n \end{bmatrix},

with components XiX_i independent and X~i\tilde{X}_i i.i.d. Gaussian such that h(Xi)=h(X~i)h(X_i) = h(\tilde{X}_i) for 1≤i≤n1 \leq i \leq n, then

h(Ax)≥h(Ax~).h(Ax) \geq h(A \tilde{x}).

The open problems asks if such an inequality is true when the random vector XX has dependent components.

A simpler version of the open problem is the following. Let X∈RNX \in \mathbb{R}^N have dependent entries and define Y=X+Z∈R2Y = X + Z \in \mathbb{R}^2 where Z∈R2Z \in \mathbb{R}^2 is noise. The open problem looks for an EPI of the form

N(Y)≥g(N(X),N(Z))N(Y) \geq g(N(X), N(Z))

for some function g(⋅,⋅)g(\cdot, \cdot). Even simpler still we could take

X=[X1X1],X = \begin{bmatrix} X_1 & X_1 \end{bmatrix},

i.e.\ X1=X2X_1 = X_2, and seek an EPI of the form in the above with N(X)N(X) replaced with N(X1)N(X_1).

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.