2
$\begingroup$

I am reading Zeros of Gaussian Analytic Functions by Mikhail Sodin and he gives an much-too-easy proof of density of zeros of a Gaussian Analytic function.

Definition A Gaussian analytic function $f(z) = \sum a_n f_n(z)$ where $a_n$ is a Gaussian normal complex random variable and $f_n(z)$ are holomorhpic functions.

Sodin studies the density of the zeros the distribution and gets a proof by Edleman-Kostlan

$$ n_f = \sum_{f(a)= 0} \delta (x-a) = \tfrac{1}{2\pi}\log | \,f \,| $$

Theorem $ \mathbb{E}[n_f]=\tfrac{1}{2\pi}\log || \,f \,|| $ where $||\, f \,||^2 = \sum |\,f_n(z)\,|^2$ the Riesz measure of the norm of $\vec{f}$.

The proof talks about the Fubini study metric and it's just three steps:

$$ \mathbb{E}[n_f] = \frac{1}{2\pi} \mathbb{E} \big[ \Delta \log |\,f\,| \big] = \frac{1}{2\pi} \log ||\,f\,||$$

The last step is an abbreviation for $\mathbb{E}\log |\vec{a}| = \log ||\vec{a}|| $

Why is this equivalent to the Crofton formula which expresses arc-length in terms of the intersection of a curve with random lines?

$$ \int_{\mathrm{Gr}_1(\mathbb{R}^2)} \# \{ \ell \cap C\} d\mu = 2\pi | C|$$

$\endgroup$

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.