A Multiscale Approach to Optimal Transport

被引:177
作者
Merigot, Quentin [1 ,2 ]
机构
[1] Univ Grenoble, Lab Jean Kuntzmann, Grenoble, France
[2] CNRS, F-75700 Paris, France
关键词
D O I
10.1111/j.1467-8659.2011.02032.x
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
In this paper, we propose an improvement of an algorithm of Aurenhammer, Hoffmann and Aronov to find a least square matching between a probability density and finite set of sites with mass constraints, in the Euclidean plane. Our algorithm exploits the multiscale nature of this optimal transport problem. We iteratively simplify the target using Lloyd's algorithm, and use the solution of the simplified problem as a rough initial solution to the more complex one. This approach allows for fast estimation of distances between measures related to optimal transport (known as Earth-mover or Wasserstein distances). We also discuss the implementation of these algorithms, and compare the original one to its multiscale counterpart.
引用
收藏
页码:1583 / 1592
页数:10
相关论文
共 23 条
[1]   Minimizing flows for the Monge-Kantorovich problem [J].
Angenent, S ;
Haker, S ;
Tannenbaum, A .
SIAM JOURNAL ON MATHEMATICAL ANALYSIS, 2003, 35 (01) :61-97
[2]  
[Anonymous], 1987, Unconstrained Optimization: Practical Methods of Optimization
[3]   Minkowski-type theorems and least-squares clustering [J].
Aurenhammer, F ;
Hoffmann, F ;
Aronov, B .
ALGORITHMICA, 1998, 20 (01) :61-76
[4]   Capacity-Constrained Point Distributions: A Variant of Lloyd's Method [J].
Balzer, Michael ;
Schloemer, Thomas ;
Deussen, Oliver .
ACM TRANSACTIONS ON GRAPHICS, 2009, 28 (03)
[5]  
Benamou JD, 2000, NUMER MATH, V84, P375, DOI 10.1007/s002119900117
[6]  
Bertsekas D. P., 1988, Annals of Operations Research, V14, P105, DOI 10.1007/BF02186476
[7]  
BOSC D, 2010, NUMERICAL APPROXIMAT
[9]  
CHAZAL F, 2010, FDN COMPUTA IN PRESS
[10]  
de Castro PMM, 2009, COMPUT GRAPH FORUM, V28, P1465