The octahedron is badly approximated by random subspaces (Q1089568)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | The octahedron is badly approximated by random subspaces |
scientific article |
Statements
The octahedron is badly approximated by random subspaces (English)
0 references
1986
0 references
Let \(B^ N_ p\) be the unit ball in \(\ell^ N_ p\) \((1\leq p\leq \infty)\). For \(X\) a Banach space, \(K\subset X\) and any subspace \(L\) of \(X\) let \(\rho_ X(K,L)=\sup \{dist(x,L):x\in K\}\) (write \(\rho_ p(K,L)\) if \(C=\ell^ N_ p)\) and \(d_ n(K,X)=\inf \{\rho_ X(K,L):L\subset \subset X,\dim L=n\}\). If \(4 \log eN\leq n<N/2\), \(1\leq p\leq 2\), then 1) \[ \int_{G_{n,N}}\rho_{\infty}(B^ N_ p,L)\,d\mu_{n,N}(L)\asymp ((1+\log (N/n))/n)^{1/2}; \] 2) there are absolute positive constants \(c_ 1,c_ 2,\theta\) such that \(\theta<1\) and \[ {\mathcal P}_{n,N}\{L\in G_{n,N}:\rho_{\infty}(B^ N_ 1,L)<c_ 1((1+\log (N/n))/n)^{1/2}\}\leq c_ 2\theta^ n, \] where \(G_{n,N}\) is the Grassmann manifold of all \(n\)-dimensional subspaces of \(\mathbb R^ N\), \(\mu_{n,N}\) the normed unitarily invariant measure on \(G_{n,N}\) and \(P_{n,N}\) the Gaussian measure on the space of all \(n\)-by-\(N\)-matrices. If \(\log N\leq n\leq N/2\), then there is an absolute positive constant \(c\) such that \[ d_ n(B^ N_ 1,\ell^ N_{\infty})\geq c((\log N)/(n \log (1+n/\log N)))^{1/2}. \]
0 references
octahedron
0 references
random subspaces
0 references
approximation of convex bodies
0 references
Grassmann manifold
0 references
Gaussian measure
0 references
0 references