Some results on approximate 1-median selection in metric spaces

  • Authors:
  • Ching-Lueh Chang

  • Affiliations:
  • -

  • Venue:
  • Theoretical Computer Science
  • Year:
  • 2012

Quantified Score

Hi-index 5.23

Visualization

Abstract

A 1-median in a finite metric space is a point with the minimum average distance to all other points. Given a positive integer n and oracle access to a distance metric on {1,2,...,n}, we study the problem of finding a 1-median. In particular, we show the nonexistence of (1) deterministic O(1)-approximation o(n)-query algorithms, (2) deterministic (2-@W(1))-approximation o(n^2)-query algorithms for graph metrics, (3) deterministic (3-@W(1))-approximation o(n^2)-query algorithms and (4) Monte-Carlo (2-@W(1))-approximation o(n)-query algorithms with an @W(1) probability of success. We also show a Monte-Carlo (2+@e)-approximation O((log^2(1/@e))/@e^3)-query algorithm with a 1-O(@e) probability of success, where @e@?(0,1).