Improving non-negative matrix factorizations through structured initialization

被引：144

作者：

Wild, S ^{[1
]}

Curry, J ^{[1
]}

Dougherty, A ^{[1
]}

机构：

[1] Univ Colorado, Dept Math Appl, Boulder, CO 80309 USA

来源：

PATTERN RECOGNITION | 2004年 / 37卷 / 11期

基金：

美国国家科学基金会;

关键词：

non-negative matrix factorization; k-means clustering; constrained optimization; rank reduction; data mining; compression; feature extraction;

D O I：

10.1016/j.patcog.2004.02.013

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

In this paper we explore a recent iterative compression technique called non-negative matrix factorization (NMF). Several special properties are obtained as a result of the constrained optimization problem of NMF. For facial images, the additive nature of NMF results in a basis of features, such as eyes, noses, and lips. We explore various methods for efficiently computing NMF, placing particular emphasis on the initialization of current algorithms. We propose using Spherical K-Means clustering to produce a structured initialization for NMF. We demonstrate some of the properties that result from this initialization and develop an efficient way of choosing the rank of the low-dimensional NMF representation. (C) 2004 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.

引用

页码：2217 / 2232

页数：16

共 18 条

[1]

BERRY M, 1999, UNDERSTANDING SEARCH

[2] The centroid decomposition: Relationships between discrete variational decompositions and SVDs [J].