Figure 3:

An illustration of the Chinese Restaurant Process. In the CRP metaphor, imagine a Chinese restaurant with an infinite number of tables. Customers (individual data entries, shown as diamonds) enter into the restaurant one by one and are seated at tables (discrete clusters, shown as circles), in the order in which they enter into the restaurant. There are parameters associated with the clusters, represented as φ_k = [μ_k, τ_k] for the cluster means and precisions. The first customer who enters into the restaurant always sits at the first table. The second customer enters and sits at the first table with probability 1/(1 + α), and the second table with probability α/(1+α), where α is a positive real number (top row, where i = 2). When the third customer enters the restaurant, he or she sits at each of the occupied table with a probability proportional to the number of previous customers already sitting there, and at the next unoccupied table with probability proportional to α.