矩阵表示的子空间学习算法综述

线性子空间学习是较为流形的一种特征提取方法,而向量表示的子空间学习方法则是目前众多线性子空间学习方法之一.近年来人们针对这一方法的缺点和不足,提出了若干矩阵表示的子空间学习方法,这些方法都是向量表示的子空间学习方法的拓展.首先介绍了几种经典的向量表示子空间学习方法,进而对几种有代表性的矩阵表示子空间学习方法进行了综述并比较详尽地给出了这些算法的推导过程,同时对算法的有效性等问题进行了分析比较,提出了一些新的理论观点和见解.

Survey on matrix representation subspace learning algorithms

Linear subspace learning techniques are very popular feature extraction algorithms. Aiming at the disadvantage of vector representation subspace learning techniques, a lot of matrix representa- tion subspace learning techniques are proposed in recent years. Matrix representation subspace learn- ing techniques are the expansion of vector representation subspace learning techniques. In this paper, several typical vector representation linear subspace learning techniques are firstly introduced. Then several representative matrix representation linear subspace learning techniques are studied. Finally, the comparison and analysis of these algorithms are given.