Cluster 3.0 is an enhanced version of Cluster, which was originally developed by Michael Eisen while at Stanford University. The main improvement consists of the k-means algorithm, which now includes multiple trials to find the best clustering solution. This is crucial for the k-means algorithm to be reliable. The routine for self-organizing maps was extended to include 2D rectangular geometries. The Euclidean distance and the city-block distance were added to the available measures of similarity.