psudo-description:
Spectral clustering is a graph-based technique for finding groups in high-dimensional data. We approach the problem of learning the similarities among data as a function of the graph laplacian, in order to optimize spectral clustering results on fuzzy and sparse data.