This is neat... I love projects like this. When the number of clusters is unknown or hard to predict ahead of time, I'd lean toward something like affinity prop

This is neat... I love projects like this. When the number of clusters is unknown or hard to predict ahead of time, I'd lean toward something like affinity propagation or mean-shift. Why k-means? http://scikit-learn.org/stable/auto_examples/cluster/plot_affinity_propagation.html http://scikit-learn.org/stable/auto_examples/cluster/plot_af... http://scikit-learn.org/stable/auto_examples/cluster/plot_mean_shift.html http://scikit-learn.org/stable/auto_examples/cluster/plot_me...