English

On constant factor approximation for earth mover distance over doubling metrics

Data Structures and Algorithms 2010-02-23 v1

Abstract

Given a metric space (X,dX)(X,d_X), the earth mover distance between two distributions over XX is defined as the minimum cost of a bipartite matching between the two distributions. The doubling dimension of a metric (X,dX)(X, d_X) is the smallest value α\alpha such that every ball in XX can be covered by 2α2^\alpha ball of half the radius. We study efficient algorithms for approximating earth mover distance over metrics with bounded doubling dimension. Given a metric (X,dX)(X, d_X), with X=n|X| = n, we can use O~(n2)\tilde O(n^2) preprocessing time to create a data structure of size O~(n1+\e)\tilde O(n^{1 + \e}), such that subsequently queried EMDs can be O(αX/\e)O(\alpha_X/\e)-approximated in O~(n)\tilde O(n) time. We also show a weaker form of sketching scheme, which we call "encoding scheme". Given (X,dX)(X, d_X), by using O~(n2)\tilde O(n^2) preprocessing time, every subsequent distribution μ\mu over XX can be encoded into F(μ)F(\mu) in O~(n1+\e)\tilde O(n^{1 + \e}) time. Given F(μ)F(\mu) and F(ν)F(\nu), the EMD between μ\mu and ν\nu can be O(αX/\e)O(\alpha_X/\e)-approximated in O~(n\e)\tilde O(n^\e) time.

Keywords

Cite

@article{arxiv.1002.4034,
  title  = {On constant factor approximation for earth mover distance over doubling metrics},
  author = {Shi Li},
  journal= {arXiv preprint arXiv:1002.4034},
  year   = {2010}
}

Comments

Extended abstract. An older version submitted to ICALP

R2 v1 2026-06-21T14:49:35.703Z