Hierarchical Dirichlet process and relative entropy

The Hierarchical Dirichlet process is a discrete random measure serving as an important prior in Bayesian non-parametrics. It motivated with the study of groups clustered data. Each group modelled through level two and all share same base distribution which itself drawn from one process. has concentration parameters at each level. main results paper are law large numbers deviations for hierarchical its mass when both converge to infinity. deviation rate functions identified explicitly. function consists terms corresponding relative entropies less than process, reflects fact that number clusters under slower growth