← All problems

Weak variance conjecture for the spectral norm of Gaussian matrices

5 years old · traced to

Let XX be a d×dd\times d centered random matrix with jointly Gaussian entries. Define its weak variance by

σ∗(X)=sup⁡v,w∈Rd∥v∥2=∥w∥2=1(E⟨Xv,w⟩2)12.\sigma_{*}(X)=\sup_{\substack{v,w\in\mathbb{R}^{d}\\\|v\|_{2}=\|w\|_{2}=1}}\left(\mathbb{E}\langle Xv,w\rangle^{2}\right)^{\frac{1}{2}}.

Weak variance conjecture. The expected spectral norm satisfies

E∥X∥≲∥E(X∗X)∥12+∥E(XX∗)∥12+log⁡d σ∗(X).\mathbb{E}\|X\|\lesssim\|\mathbb{E}(X^{*}X)\|^{\frac{1}{2}}+\|\mathbb{E}(XX^{*})\|^{\frac{1}{2}}+\sqrt{\log d}\,\sigma_{*}(X).

The conjecture proposes that the weak variance is the parameter governing whether a logarithmic factor is necessary in noncommutative Khintchine-type bounds.

References

Status

  • Solved: Afonso S. Bandeira, March T. Boedihardjo, and Ramon van Handel. Matrix concentration inequalities and free probability (arXiv, 2021) https://arxiv.org/abs/2108.06312

Source: Afonso S. Bandeira and March T. Boedihardjo, “The spectral norm of Gaussian matrices with correlated entries”, arXiv:2104.02662 (2021).

Progress summary

Never refreshed

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Solutions 0

No solutions have been posted yet.