Affinity Propagation, and other Data Clustering Techniques

In our research we sought to create implementations of several common clustering algorithms and a relatively new approach called Affinity Propagation. Our objective was to compare the techniques by running tests on one and two dimensional datasets provided by Professor Trono. Dave Kronenberg implemented a standard randomly seeded K-Means clustering program and many related support functions. We also implemented a program to run three variations of Graph Linkage clustering: Single-Link, Complete-Link, and Centroid-Link, and we incorporated the Silhouette Index method of cluster verification into this program to rate the linkage strategies and to choose an optimal clustering (k-value) from the generated dendrogram. Finally, we completed an Affinity Propagation program which closely follows the details laid out by Frey and Dueck [1]. Using these three programs we clustered a variety of one and two dimensional datasets, and we analyzed the results to confirm that Affinity Propagation produces clustering errors competitive with that of K-Means in an order of magnitude less time.