User:Andrea wind/PCCA+

PCCA+ (Robust Perron Cluster Analysis)
PCCA+ is a spectral clustering technique in multivariate statistics used for the clustering of sets of data points. In the original articles, PCCA+ make use of the spectrum (eigenvalues and eigenvectors) of the random walk normalized Laplacian of a similarity matrix, which consists of a quantitative assessment of the relative similarity of each pair of points in the given dataset. In recent articles, it is proposed to use a Schur decomposition of a modified similarity matrix, in order to account for non-symmetric relative similarities.

In contrast to standard spectral clustering methods, PCCA+ generates one membership vector for each cluster. Thus, the data points can belong to more than one cluster with a different level of membership. The membership vectors have non-negative entries between 0 and 1. Therefore, PCCA+ can also be seen as a special method in Fuzzy Clustering.