ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let XX have prior distribution P0P_0, and consider the rank-one symmetric matrix estimation model

Yi,j=λnXiXj+Zi,j,Y_{i,j}=\sqrt{\frac{\lambda}{n}}X_iX_j+Z_{i,j},

where Zi,jZ_{i,j} are standard Gaussian noises. Assume EP0[X]=0\mathbb{E}_{P_0}[X]=0 and EP0[X2]=1\mathbb{E}_{P_0}[X^2]=1. Let λc\lambda_c be the critical value defined by the model's fixed-point criterion, and let MSE⁡\operatorname{MSE} denote mean square error and DMSE⁡\operatorname{DMSE} the error of the dummy estimator. The hard-but-detectable conjecture. If λc<1\lambda_c<1, then achieving an MSE⁡\operatorname{MSE} better than the dummy estimator is computationally hard for λ∈(λc,1)\lambda\in(\lambda_c,1). This conjecture extends the hard-but-detectable phenomenon from sparse PCA and related community-detection models; the stated proposition establishes information-theoretic achievability in this interval, but does not establish efficient achievability or computational hardness.

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.