ProbXiv
Explorersign in
Problem archiveProblem record

Statement

In the Gaussian sequence model, observe independent vectors Yi=θi+σξi∈RpY_i=\theta_i+\sigma\xi_i\in\mathbb{R}^p, i=1,…,ni=1,\ldots,n, where ξi∼iidNp(0,Ip)\xi_i\stackrel{\mathrm{iid}}\sim N_p(0,I_p), σ\sigma is known, and the matrix Θ=[θ1,…,θn]∈Rp×n\Theta=[\theta_1,\ldots,\theta_n]\in\mathbb{R}^{p\times n} has ss nonzero columns. For L(Θ)=∑i∈[n]θiL(\Theta)=\sum_{i\in[n]}\theta_i and rmmx(n,p,s)=inf⁡L^sup⁡ΘE[∥L^−L(Θ)∥22]r_{\mathrm{mmx}}(n,p,s)=\inf_{\widehat L}\sup_{\Theta}\mathbb{E}[\|\widehat L-L(\Theta)\|_2^2], determine the nonasymptotic minimax rate, up to possible logarithmic factors, in the large-sparsity regime described in the report as s≥const(n∨p)s\ge \mathrm{const}(\sqrt n\vee p). The report only establishes that this rate lies between n+spn+sp and (s2∧np)+sp(s^2\wedge np)+sp in that regime.

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.