随机增量张量奇异值分解与人脸识别新算法

A Randomized Tensor Singular Value Decomposition with Increament for Face Recognition

本文新提出随机增量张量奇异值分解方法.当数据逐步增加时,新方法能够在保持原数据的随机奇异值分解基础上,通过计算新增数据的奇异值分解得到更新后数据的张量奇异值分解.基于随机增量张量奇异值分解建立新的人脸识别模型.数值实验表明新模型与已有人脸识别模型相比具有较高的识别率.

In this paper,the randomized tensor singular value decomposition with increment is proposed.In the process of updating the training datasets,the existing results of the original datasets can be maintained and utilized to obtain the SVD decomposition of the new datasets.Based on the randomized tensor singular value decomposition with increment,a new model of face recognition is established.Numerical experiments show that the new model has higher recognition rate than existing face recognition models.