Refining a $k$-nearest neighbor graph for a computationally efficient spectral clustering

Open in new window