一种基于加速迭代的大数据集谱聚类方法

Spectral Clustering Algorithm for Large Scale Data Set Based on Accelerating Iterative Method

传统谱聚类算法的诸多优点只适合小数据集。根据Laplacian矩阵的特点重新构造新的Gram矩阵,输入新构造矩阵的若干列,然后利用加速迭代法解决大数据集的谱聚类特征提取问题,使得在大数据集条件下,谱聚类算法只需要很小的空间复杂度就可达到非常快的计算速度。

The advantage of the traditional spectral clustering algorithm is applicable in the small scale data set.A new method was proposed in the light of the laplacian matrix characteristics.First,a new Gram matrix was reconstructed and some lies of the new matrix were needed,then the eigen-decomposition based on accelerating iterative method was solved.The calculation speed of the proposed method is very fast and the space complexity is small for large scale data set.