Magnitude Sensitive Competitive Learning

被引：4

作者：

Pelayo, Enrique ^{[1
]}

Buldain, David ^{[1
]}

Orrite, Carlos ^{[1
]}

机构：

[1] Univ Zaragoza, Aragon Inst Engn Res, CVLab, E-50009 Zaragoza, Spain

来源：

NEUROCOMPUTING | 2013年 / 112卷

关键词：

Competitive learning; Neural Networks; Vector quantization; Surface modelling; Color quantization; Interpolation; SELF-ORGANIZING MAPS; QUANTIZATION; ATTENTION; INFANTS;

D O I：

10.1016/j.neucom.2012.11.039

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

This paper presents a new neural method for unsupervised learning denoted as Magnitude Sensitive Competitive Learning (MSCL), which has the property of distributing the unit centroids following any magnitude calculated from the unit parameters or the input data inside its Voronoi region. This controlled behavior permits it to outperform standard Competitive Learning algorithms that only tend to concentrate neurons according to the input data density, when other kind of data information processing is desired. Some examples applying different target functions show the MSCL possibilities in several applications as data-series interpolation, surface modelling from 3D point clouds and color quantization (CQ). (C) 2013 Elsevier B.V. All rights reserved.

引用

页码：4 / 18

页数：15

共 39 条

[1] COMPETITIVE LEARNING ALGORITHMS FOR VECTOR QUANTIZATION [J].