摘要
作为非负矩阵分解的多线性推广,非负张量分解已被成功地应用在信号处理、计算机视觉、数据挖掘和神经科学等领域中。提出了非负张量分解的一种快速算法。首先,将大的张量数据视做多元连续函数的离散化,对其进行采样得到一个小张量;其次,对小张量执行非负分解,可得到它的重构张量;然后,对于采样后的重构张量,使用二维线性插值方法对原始张量进行重构;最后,实验结果表明快速张量分解算法的有效性。
As the multi-linear extension of nonnegative matrix factorization, nonnegative tensor factorization has been successfully applied in many fields including signal processing, computer vision, data mining and neuroscience. This paper proposed a fast algorithm to nonnegative tensor factorization. Firstly, regarded a lager tensor data as the discretization of multivariate continuous function and obtained a corresponding smaller tensor data by sampling. Secondly, performed the nonnegative factorization on the small tensor and easily computed the corresponding reconstruction tensor. Then, employed for the above reconstruction tensor, two-dimensional linear interpolation to reconstruct the original tensor. Finally,the experimental results show the effectiveness of the proposed fast algorithm to nonnegative tensor faetorization.
出处
《计算机应用研究》
CSCD
北大核心
2011年第12期4475-4477,共3页
Application Research of Computers
基金
陕西省自然科学基金资助项目(JQ1003)
陕西省教育厅专项科研计划资助项目(09JK545)
关键词
非负张量分解
非负矩阵分解
快速算法
采样
插值
重构
nonnegative tensor factorization
nonnegative matrix factorization
fast algorithm
sampling
interpolation
reconstruction
