A mixed iteration for nonnegative matrix factorizations

被引：3

作者：

Soltuz, Stefan M. ^{[1
,2
]}

Rhoades, B. E. ^{[3
]}

机构：

[1] Dawson Coll, Dept Math, Montreal, PQ H3Z 1A4, Canada

[2] Tiberiu Popoviciu Inst Numer Anal, Cluj Napoca, Romania

[3] Indiana Univ, Dept Math, Bloomingtron, IN USA

来源：

APPLIED MATHEMATICS AND COMPUTATION | 2013年 / 219卷 / 18期

关键词：

Non-negative matrix factorization; Lee-Seung iteration; ALGORITHMS;

D O I：

10.1016/j.amc.2013.03.124

中图分类号：

O29 [应用数学];

学科分类号：

070104 ;

摘要：

We show that, under appropriate conditions, one can create a hybrid between two given iterations which can perform better than either of the original ones. This fact provides a freedom of choice. We also give numerical examples in which we compare our hybrid with the dedicated Lee-Seung iteration. (C) 2013 Elsevier Inc. All rights reserved.

引用

页码：9847 / 9855

页数：9

共 19 条

[11]

Lee DD, 2001, ADV NEUR IN, V13, P556

[12] On the convergence of multiplicative update algorithms for nonnegative matrix factorization [J].

Lin, Chih-Jen .

IEEE TRANSACTIONS ON NEURAL NETWORKS, 2007, 18 (06) :1589-1596

[13]

LIN CJ, 2007, NEURAL COMP IN PRESS

[14]

Schmidt M., 2008, THESIS TU DENMARK

[15]

Soltuz Stefan M., 2009, WORKSH 15 STAT SIGN

[16]

Soltuz Stefan M., 2001, REV ANAL NUMERIQUE T, V29, P207

[17]

Strang G., 1986, Introduction to Applied Mathematics

[18]

Wang W., 2008, NONNEGATIVE MATRIX F

[19]

Zdenuk R., 2007, NEUROCOMPUTING, V71, P2309

← 1 2 →