Feature selection for multiclass discrimination via mixed-integer linear programming

被引：29
|
作者：
Iannarilli, FJ
Rubin, PA
机构：
[1] Aerodyne Res Inc, Billerica, MA 01821 USA
[2] Michigan State Univ, Eli Broad Grad Sch Management, Dept Management, E Lansing, MI 48824 USA
来源：
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE | 2003年 / 25卷 / 06期
关键词：
feature selection; discrimination; classification; mixed-integer; linear programming; branch-and-bound;
D O I：
10.1109/TPAMI.2003.1201827
中图分类号：
TP18 [人工智能理论];
学科分类号：
081104 ; 0812 ; 0835 ; 1405 ;
摘要：
We reformulate branch-and-bound feature selection employing L-infinity or particular L-p metrics, as mixed-Integer linear programming (MILP) problems, affording convenience of widely available MILP solvers. These formulations offer direct influence over individual pairwise interclass margins, which is useful for feature selection in multiclass settings.
引用
收藏
页码：779 / 783
页数：5
相关论文
共 50 条
12345