ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Part (3) of the problem: Let s≤d≤ms\leq d\leq m and z1,…,zm∈Rdz_1,\dots,z_m\in \mathbb{R}^d i.i.d. random vectors with i.i.d. entries (zk)j={−1s with probability s2m0 with probability 1−sm1s with probability s2m\left( z_k\right)_j = \left\{ \begin{array}{rcc} -\frac1{\sqrt{s}} & \text{ with probability } & \frac{s}{2m} \\ 0 & \text{ with probability } & 1-\frac{s}{m} \\ \frac1{\sqrt{s}} & \text{ with probability } & \frac{s}{2m} \end{array} \right. Note that EzkzkT=1mId×d.\mathbb{E} z_kz_k^T = \frac1m I_{d\times d}. The conjecture is that, there exists c1c_1 and c2c_2 positive universal constants such that Prob{∥∑k=1m[zkzkT−EzkzkT]∥≥ε}<δ,\mathrm{Prob}\left\{ \left\| \sum_{k=1}^m \left[z_k z_k^T - \mathbb{E} z_k z_k^T\right] \right\| \geq \varepsilon \right\} < \delta, for m≥c1d+log⁡(1δ)ε2m\geq c_1 \frac{d+\log\left( \frac1{\delta} \right)}{\varepsilon^2} and s≥c2log⁡(dδ)ε2s\geq c_2 \frac{\log\left( \frac{d}{\delta} \right)}{\varepsilon^2}.

Record

Added
Links

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.