ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let β∈[−1,∞)\beta \in [-1,\infty) and γ>0\gamma>0. For each n∈Nn\in\mathbb N, set N=⌈n/γ⌉N=\lceil n/\gamma\rceil, and let P\mathbb P and Q\mathbb Q be the spiked Wishart distributions defined by the null model Q\mathbb Q with independent samples yi∼N(0,In)\bm y_i\sim\mathcal N(\bm 0,\bm I_n) and the planted model P\mathbb P obtained by sampling x∼Unif⁡({±1}n)\bm x\sim\operatorname{Unif}(\{\pm1\}^n) and then independently sampling yi∼N(0,In+(β/n)xx⊤)\bm y_i\sim\mathcal N(\bm 0,\bm I_n+(\beta/n)\bm x\bm x^{\top}). Write this sequence of distribution pairs as Wishart(β,γ)\mathsf{Wishart}(\beta,\gamma). Hardness of the Wishart model. If β>−1\beta>-1 and β2<γ\beta^2<\gamma, there is no polynomial-time hypothesis test between the distributions of Wishart(β,γ)\mathsf{Wishart}(\beta,\gamma): no randomized algorithm f:(Rn)N→{p,q}f:(\mathbb R^n)^N\to\{\mathtt p,\mathtt q\} running in polynomial time in nn satisfies

lim⁡n→∞P[f(y1,…,yN)=p]=lim⁡n→∞Q[f(y1,…,yN)=q]=1.\lim_{n\to\infty}\mathbb P[f(\bm y_1,\dots,\bm y_N)=\mathtt p]=\lim_{n\to\infty}\mathbb Q[f(\bm y_1,\dots,\bm y_N)=\mathtt q]=1.

This conjecture is the key computational-hardness assumption used to obtain evidence that the constant 11 in spectral conditions for general-purpose Ising-model sampling and free-energy approximation is optimal; its status is not resolved in the source.

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.