APPROACH TO PERFORMANCE OPTIMIZATION IN FREQUENCY-WEIGHTED MEMORY NETWORK PATTERN CLASSIFIERS

被引：8

作者：

FAIRHURST, MC

WAHAB, HMSA

机构：

来源：

ELECTRONICS LETTERS | 1987年 / 23卷 / 21期

关键词：

DATA STORAGE; DIGITAL;

D O I：

10.1049/el:19870779

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

The performance of memory network pattern classifiers based on a feature frequency-weighting scheme can suffer from an inappropriate arbitrary choice of zero-frequency component in the computation of discriminant functions. This letter assesses the effect of optimizing this factor, emphasising the value of optimization in pairwise class discrimination, which is of particular relevance in the design of efficient hierarchical classifer architectures.

引用

页码：1116 / 1118

页数：3

共 10 条

[1]

Batchelor BG, 1978, PATTERN RECOGNITION

[2]

Devijver P.A., 1982, PATTERN RECOGNITION

[3] A two-layer memory network architecture for a pattern classifier [J].