Let $X\sim N(\mu,\sigma^2I)\in \mathbb{R}^n$ be a non-centered ($\mu\neq 0$) Gaussian vector with independent coordinates. I'm wondering if there is any sharp lower bound of the following probability: $$\mathbb{P}\left(X^TAX - \mathbb{E}(X^TAX) \geq \delta\right)$$ where $A$ is a symmetric matrix and $\delta>0$.

I know that when $X$ is centered, the upper bound is well-known, which is the Hanson-Wright inequality. However, in my case, $X$ is not centered and I'm interested in the lower bound.

  • $\begingroup$ It seems like you are interested in anti-concentration inequalities, or small ball probabilities. This paper might be useful: arxiv.org/abs/1507.00829. $\endgroup$ – Ankitp Feb 15 '19 at 13:28

Yeah you are looking for the anti concentration inequality for Gaussian polynomials. For degree $d$ polynomials the probability that you lie in an $\epsilon$ interval is upper bounded by $O(d\epsilon^{1/d})$. Check out Carbery Wright 2001.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.