L1-范数子空间技术的鲁棒建模综述

子空间学习可以通过多种技术来开展,对一些流行且被广泛使用的子空间学习方法,简要回顾其基于L1-范数的鲁棒建模.从主成分分析开始介绍子空间学习技术、线性判别分析以及更一般的图嵌入框架.作为L1-范数的综合利用,进一步讨论具有稀疏性的鲁棒建模.此外,还论述一些应用在神经科学中的相关子空间学习技术.最后,针对基于L1-范数的子空间学习的求解问题,介绍一个有力工具,即边界优化技术.

L1-norm based subspace techniques for robust modelling: a brief review

Subspace learning can be performed with plenty of techniques.The paper was to briefly review L1-norm based robust modelling for the most popular and widely used subspace learning methods.We commenced the subspace learning technique from principal component analysis,and proceeded to linear discriminant analysis and then the more general framework called graph embedding.As a comprehensive utilization of the L1-norm,robust modelling plus sparsity was further discussed.Besides,some relative subspace learning techniques applied in neuroscience were gone through.Finally,the bound optimization technique,a useful tool for tackling the L1-norm based subspace learning,was described.