Using diversity measures for generating error-correcting output codes in classifier ensembles

被引：50

作者：

Kuncheva, LI ^{[1
]}

机构：

[1] Univ Wales, Sch Informat, Bangor LL57 1UT, Gwynedd, Wales

来源：

PATTERN RECOGNITION LETTERS | 2005年 / 26卷 / 01期

关键词：

statistical pattern recognition; classifier ensembles; diversity measures; error-correcting output codes (ECOC); evolutionary algorithms;

D O I：

10.1016/j.patrec.2004.08.019

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Error-correcting output codes (ECOC) are used to design diverse classifier ensembles. Diversity within ECOC is traditionally measured by Hamming distance. Here we argue that this measure is insufficient for assessing the quality of code for the purposes of building accurate ensembles. We propose to use diversity measures from the literature on classifier ensembles and suggest an evolutionary algorithm to construct the code. (C) 2004 Elsevier B.V. All rights reserved.

引用

页码：83 / 90

页数：8

共 9 条

[1]

Aha DW, 1997, AI APPLICATIONS, V11, P13

[2]

DIETTERICH TG, 1991, PROCEEDINGS : NINTH NATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOLS 1 AND 2, P572

[3]

Dietterich TG, 1994, J ARTIF INTELL RES, V2, P263

[4] Measures of diversity in classifier ensembles and their relationship with the ensemble accuracy [J].

Kuncheva, LI ;

Whitaker, CJ .

MACHINE LEARNING, 2003, 51 (02) :181-207

[5]

Masulli F, 2000, KES'2000: FOURTH INTERNATIONAL CONFERENCE ON KNOWLEDGE-BASED INTELLIGENT ENGINEERING SYSTEMS & ALLIED TECHNOLOGIES, VOLS 1 AND 2, PROCEEDINGS, P788, DOI 10.1109/KES.2000.884164

[6]

Masulli F, 2000, LECT NOTES COMPUT SC, V1857, P107

[7]

Shapire R. E., 1997, P 14 INT C MACH LEAR

[8]

Windeatt T., 2003, Information Fusion, V4, P11, DOI 10.1016/S1566-2535(02)00101-X

[9]

Windeatt T., 2001, Information Fusion, V2, P103, DOI 10.1016/S1566-2535(01)00029-X

← 1 →