Ars Technica: As data sets grow ever larger, better algorithms are being developed to help parse them. One common approach is cluster analysis, whereby data are grouped in clusters, based on having similar characteristics. Two such algorithms, k-means and k-medoids, are good at detecting spherical clusters, while methods that use density-based clustering of applications with noise are better at detecting nonspherical clusters. Now researchers have combined features of those different algorithms to come up with one that can do a fast search and find of density peaks and identify the correct number of clusters. Such an algorithm could be put to good use in the fields of bioinformatics and genomics because of the amount of data being generated in those areas.
