Figure 4. Characteristics of sampled partitions.

We calculate how often each pair of users (A) or movies (B) are co-classified in the same group in the sampled partitions. (A) Probability that a pair of users have the same gender (circles), and their age difference (squares), as a function of their co-classification frequency. (B) Overlap between the genres of a pair of movies (see Methods) as a function of their co-classification frequency.