Spectral clustering based on iterative optimization for large-scale and high-dimensional data

Spectral graph theoretic methods have been a fundamental and important topic in the field of manifold learning and it has become a vital tool in data clustering. However, spectral clustering approaches are limited by their computational demands. It would be too expensive to provide an optimal approximation for spectral decomposition in dealing with large-scale and high-dimensional data sets. On the other hand, the rapid development of data on the Web has posed many rising challenges to the traditional single-task clustering, while the multi-task clustering provides many new thoughts for real-world applications such as video segmentation. In this paper, we will study a Spectral Clustering based on Iterative Optimization (SCIO), which solves the spectral decomposition problem of large-scale and high-dimensional data sets and it well performs on multi-task clustering. Extensive experiments on various synthetic data sets and real-world data sets demonstrate that the proposed method provides an efficient solution for spectral clustering.