Consider the Haar measure $\mu$ on the Grassmannian $G(n, k)$ of $k$-dimensional subspaces in $\mathbb{R}^n.$ Now, pick pairs of subspaces uniformly at random with respect to $\mu,$ and compute their Grassmann geodesic distance (this is equal to $\|\mathbf{\theta}\|,$ where $\mathbf{\theta}$ is the vector of principal angles between the subspaces. I would like a formula (or an asymptotic expression for large $n$) for the mean of this distance.

Experiments show that this quantity seems to be normal-ish, and that the variance goes to zero - the latter is not a surprise, and is due to Gromov-Milman 1983.

  • 1
    $\begingroup$ Probably a silly remark, but for a fixed $k$ and large $n$, doesn't it tend to the distance between $\mathrm{span}\{e_1,\dots,e_k\}$ and $\mathrm{span}\{e_{k+1},\dots,e_{2k}\}$, where $e_1,\dots,e_{2k}$ is an orthonormal basis in $\mathbb{R}^{2k}$? You can sample two subspaces as a span of $k$ independent vectors each and these will be nearly orthogonal for large $n$. $\endgroup$ – Kostya_I Jan 4 at 11:07

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.