Combinatorial Preconditioners for Proximal Algorithms on Graphs

被引:0
|
作者
Moellenhoff, Thomas [1 ]
Ye, Zhenzhang [1 ]
Wu, Tao [1 ]
Cremers, Daniel [1 ]
机构
[1] Tech Univ Munich, Dept Comp Sci, Munich, Germany
来源
INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 84 | 2018年 / 84卷
关键词
OPTIMIZATION;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We present a novel preconditioning technique for proximal optimization methods that relies on graph algorithms to construct effective preconditioners. Such combinatorial preconditioners arise from partitioning the graph into forests. We prove that certain decompositions lead to a theoretically optimal condition number. We also show how ideal decompositions can be realized using matroid partitioning and propose efficient greedy variants thereof for large-scale problems. Coupled with specialized solvers for the resulting scaled proximal subproblems, the preconditioned algorithm achieves competitive performance in machine learning and vision applications.
引用
收藏
页数:10
相关论文
共 50 条