谱聚类算法建立在谱图理论基础上，与传统的聚类算法相比，它具有能在任意形状的样本空间上聚类且收敛于全局最优解的优点。



该算法首先根据给定的样本数据集定义一个描述成对数据点相似度的亲合矩阵,并且计算矩阵的特征值和特征向量 ， 然后选择合适 的特征向量聚类不同的数据点。谱聚类算法最初用于计算机视觉 、VLS I 设计等领域， 最近才开始用于机器学习中，并迅速成为国际上机器学习领域的研究热点。-Spectral clustering algorithm based on the reproduction diagram theory, compared with the traditional clustering algorithm, it can have on the arbitrary shape of sample space clustering, and the advantage of converge to global optimal solution.The algorithm firstly according to the given sample data set defines a descr iption of the data points of similarity in pairs of affinity matrix, and computing the matrix eigenvalue and characteristic vector, and then choose the appropriate feature vector clustering of different data points.Spectral clustering algorithm was originally used in computer vision, VLS in areas such as design, I recently started used in machine learning, and quickly became a hot issue in the field of international machine learning research.