Bounds on Dimension Reduction in the Nuclear Norm

From MaRDI portal
Publication:5115972

DOI10.1007/978-3-030-46762-3_13zbMATH Open1452.46017arXiv1901.09480OpenAlexW2914148868MaRDI QIDQ5115972FDOQ5115972

Thomas Vidick, Oded Regev

Publication date: 21 August 2020

Published in: Lecture Notes in Mathematics (Search for Journal in Brave)

Abstract: ewcommandschsscriptstylemathsfS1For all nge1, we give an explicit construction of mimesm matrices A1,ldots,An with m=2lfloorn/2floor such that for any d and dimesd matrices A'1,ldots,A'n that satisfy [ |A'_i-A'_j|_{schs} ,leq, |A_i-A_j|_{schs},leq, (1+delta) |A'_i-A'_j|_{schs} ] for all i,jin1,ldots,n and small enough delta=O(nc), where c>0 is a universal constant, it must be the case that dge2lfloorn/2floor1. This stands in contrast to the metric theory of commutative ellp spaces, as it is known that for any pgeq1, any n points in ellp embed exactly in ellpd for d=n(n1)/2. Our proof is based on matrices derived from a representation of the Clifford algebra generated by n anti-commuting Hermitian matrices that square to identity, and borrows ideas from the analysis of nonlocal games in quantum information theory.


Full work available at URL: https://arxiv.org/abs/1901.09480




Recommendations




Cites Work


Cited In (4)





This page was built for publication: Bounds on Dimension Reduction in the Nuclear Norm

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5115972)