前言
题目里有个latent，其实就是，不在原始特征上做子空间聚类，而是在寻找latent representation的时候，同时基于这个representation进行data reconstruction.
至于multiview，我感觉是用来凑数的吧
intro
Recently, the subspace clustering based on selfrepresentation has been proposed, where each data point can be expressed with a linear combination of the data points themselves.
而multiview的初衷是，每一个点都可以described with information from multiple sources of features、
算法
就是将self representation和multiew的对于目标函数组合一起，再搞点权值。
不细看了。