广义Bregman距离NTF算法及其在人脸识别中的应用

NTF algorithm based on generalized Bregman distance and its application in face recognition

柏锦萱 ¹王君 ¹徐常青¹

扫码查看

作者信息

1. 苏州科技大学数学科学学院,江苏苏州 215009
折叠

摘要

非负张量分解(Nonnegative Tensor Factorization,NTF)方法将一个元素非负的张量表为一些秩-1的非负张量之和,非负矩阵分解(Nonnegative Matrix Factorization,NMF)是NTF在矩阵情形下的特殊情形.首先,介绍基于Kullback-Leibler(KL)散度和Bregman距离下的非负矩阵分解;然后,给出了基于KL散度和Bregman距离的非负张量分解算法;最后,将NMF和NTF用于人脸识别的特征提取.结果表明,NTF方法优于其他方法.

Abstract

Non-negative tensor factorization(NTF)is a technique used to decompose a non-negative tensor into the sum of some non-negative rank-one tensors.It is an extension of non-negative matrix factorization(NMF)to higher-order tensors.We introduced the non-negative matrix factorization based on Kullback-Leibler(KL)di-vergence and Bregman distance,and provided the NTF algorithm based on KL divergence and Bregman distance.Finally,NMF and NTF were used for the feature extraction in the face recognition.The results show that NTF outperforms other methods.

关键词

Bregman距离/Kullback-Leibler散度/面部识别/非负张量分解/非负矩阵分解

Key words

Bregman distance/Kullback-Leibler divergence/face recognition/non-negative tensor factorization/non-negative matrix factorization

引用本文复制引用

出版年

2024

苏州科技大学学报(自然科学版)

苏州科技学院

苏州科技大学学报(自然科学版)

影响因子：0.185

ISSN：2096-3289

段落导航