Recently, the Riemannian manifold has received special attention in unsupervised clustering since the real-world visual data usually resides on a special manifold where Euclidean geometry fails to capture. Although many clustering algorithms have been proposed, most of them use only a single geometric model to describe the data. In this paper, a multi-geometric subspace clustering model is proposed, and the subspace representation is learned together by constructing a shared affinity matrix of multi-order data. Experimental results on several different types of datasets show that the clustering performance of our proposed algorithm is better than most of subspaces algorithms.

中文翻译：

---

多几何稀疏子空间聚类

最近，黎曼流形在无监督聚类中受到了特别的关注，因为现实世界中的视觉数据通常位于欧氏几何无法捕获的特殊流形上。尽管已提出了许多聚类算法，但是大多数聚类算法仅使用单个几何模型来描述数据。本文提出了一种多几何子空间聚类模型，并通过构造一个共享的多阶数据亲和矩阵来学习子空间表示。在几种不同类型的数据集上的实验结果表明，我们提出的算法的聚类性能优于大多数子空间算法。