Expected distance from the span

60 Views Asked by At

I would like to pick at random $3$ vectors on the sphere $\mathbb{S}^2$ in $\mathbb{R}^3$ and compute the expected distance of one vector to the span of the others. More precisely: let $a_1$, $a_2$, $a_3 \in \mathbb{S}^2$ be picked uniformly. What can we expect of $\underset{i,j,k\in\{1,2,3\} i\neq j \neq k}{min}dist(a_i, span(a_j,a_k))$? (I am also interested in the higher dimensional analogue)

My approach to this was a bit naïve: two vectors in $\mathbb{S}^2$ define a hyperplane through the origin, I calculate the measure of the set $\{-d \leq e_1 \leq d\}^c\cap \mathbb{S}^2$ and divide it by the total measure of $\mathbb{S}^2$, call it $g$. So given, say $a_1, a_2$, the probability that $a_3$ is closer than $d$ to the span of the other vectors is $1-g$. To calculate the minimum over all 3 combinations, I just use the union bound - so 3(1-g) is the probability the minimum is greater than $d$ (since the respective distances to the span are not independent (?) ). Is there a more refined way to do so (keeping in mind that I would like to use it for bigger dimensions than 3). How to detect the dependencies?

Any suggestion or reference is very much appreciated!

1

There are 1 best solutions below

1
On

Without loss of generality, let the span of the first two vectors be the $z$ plane. The sphere $\mathbb S^2$ has the special property that the coordinates of uniformly random points on it are uniformly distributed. Thus the expected distance is the expected absolute value of the uniform distribution on $[-1,1]$, which is $\frac12$.