子空间聚类的重建模型及其快速算法

有限样本的子空间数据聚类建模及其大规模计算是子空间学习面临的主要问题.现有的大多数模型都不适合大规模计算.本文提出了一个新的优化模型,结合谱投影反馈和辅助信息优化.在提升模型的学习能力的同时,采用高效的分片符号更新算法,可以适合大规模计算.我们用较大规模的模拟例子和实际例子,分析检验了新的优化模型及其快速算法的优于现有其他模型与算法的有效性.

RECONSTRUCTION MODEL AND FAST ALGORITHM FOR SUBSPACE CLUSTERING

Effective models and scalable algorithms are two key issues of subspace learning from limited samples. State-of-the-art methods usually suffer from their rapidly increasing costs when the data scale becomes large. This paper proposes a new reconstruction model that utilizes spectral projection and auxiliary information from samples for improving its capacity for subspace learning. A fast algorithm based on active piece-wise sign updating is then proposed for efficiently solving the problem in large scale. Comprehensive experiments on both synthetic and real-world datasets are reported to demonstrate the effectiveness and efficiency of the proposed model and the performance better than the existing algorithms for subspace learning.