NNS lower bounds via metric expansion for l

  • Authors:
  • Michael Kapralov;Rina Panigrahy

  • Affiliations:
  • Stanford iCME, Stanford, CA;MSR Silicon Valley, Mountain View, CA

  • Venue:
  • ICALP'12 Proceedings of the 39th international colloquium conference on Automata, Languages, and Programming - Volume Part I
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

We give new lower bounds for randomized NNS data structures in the cell probe model based on robust metric expansion for two metric spaces: l∞ and Earth Mover Distance (EMD) in high dimensions. In particular, our results imply stronger non-embedability for these metric spaces into l1. The main components of our approach are a strengthening of the isoperimetric inequality for the distribution on l∞ introduced by Andoni et al [FOCS'08] and a robust isoperimetric inequality for EMD on quotients of the boolean hypercube.