1
$\begingroup$

Let $\mathbf{p}$ be a point on a unit $n$-sphere $S_n$ and let $\mathcal{H}_n$ be the set all hyperplanes passing through the center of $S_n$.


Question: What is the simplest way to calculate the expected Euclidean distance $\mathbb{E}[d]$ between $\mathbf{p}$ and a hyperplane selected uniformly at random from $\mathcal {H}_n$?



Should we calculate

$$\mathbb{E}[d]=\mathbb{E}\left[\frac{|x_n|}{\sqrt{\sum_{i=1}^n x_i^2}}\right]$$

where $x_1, x_2, \ldots, x_n$ are the values taken by the i.i.d. random variables $X_1, X_2, \ldots, X_n \sim \mathcal{N}(0,1)$?

$\endgroup$
3
1
$\begingroup$

Your method would yield the right answer, but it's more complicated than it needs to be. From this answer, to generate a random hyperplane passing through the origin, first pick a random point from the surface of the $n$-sphere. The hyperplane would then be defined by $\langle x,u\rangle=0$.

WLOG, let $\mathbf{p} = (1, \underbrace{0, ..., 0}_{n-1})$. The distance from the hyperplane defined by $\langle x,u\rangle=0$ to $\mathbf{p}$ would then be $\left| \langle u, \mathbf{p}\rangle \right|$. This is then simply $|u_1|$. This is the integral of $|x_1|$ over the $n$-sphere, divided by the "surface area" of the sphere. From this, this is given by $$\frac{\frac{2\pi^{\frac{n-1}{2}}}{\Gamma\left(\frac{n+1}{2}\right)}}{\frac{2\pi^{\frac{n}{2}}}{\Gamma\left(\frac{n}{2}\right)}} = \frac{\Gamma\left( \frac{n}{2}-1 \right)}{\Gamma\left(\frac{n-1}{2} \right)\sqrt{\pi}}$$

$\endgroup$
1
  • $\begingroup$ Thank you @VarunVejalla ! $\endgroup$ Nov 27 '20 at 1:52

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.