Lipschitz normal embeddings in the space of matrices (Q667625): Difference between revisions

A semi-algebraic subset in $\mathbb{R}^n$ or $\mathbb{C}^n$ is called Lipschitz normally embedded if the inner metric and the outer metric are bilipschitz equivalent. The authors prove Lipschitz normal embeddedness for the connected components of the space $X_r$ of matrices of rank $r$ inside the space of rectangular matrices and also inside the space of (skew-)symmetric or hermitian square matrices, and Lipschitz normal embeddedness of the closure $\overline{X}_r$, with explicit bilipschitz constants. The proof is based on normal forms, giving a local description of the spaces involved. They show that for a linear subspace $V$ intersecting all $X_r$ transversally for all $r$ also the intersection $V\cap \overline{X}_r$ is Lipschitz normally embedded and give examples showing that this is no longer true under more general assumptions. They also discuss the case of determinantal varieties. An essentially isolated determinantal singularity (EIDS) defined by a matrix with linear entries is globally Lipschitz normally embedded, but such a result cannot be expected for general EIDS.