Is affinity propagation only good at finding a large number of quite small clusters?
It depends on what you mean by “large” and “small”. For example, it beats other methods at finding 100 clusters in 17,770 Netflix movies. While “100” may seem like a large number of clusters, it is not unreasonable to think that there may be 100 distinct types of movies. Also, on average there are 178 points per cluster, which is not “quite small”. However, if you’re looking for just a few clusters (eg, 1 to 5), you’d probably be better off using a simple method (see below).
