"Bayesian Nonparametrics "(Springer)

巻頭言より。

To sum up, Bayesian nonparametrics is sufficiently well developed to take care of many problems. Computation of the posterior is numerically feasible for several class of priors. We now know a fair amount of asymptotic behaviour of posteriors for different priors to ensure consistency at plausible Base measures.

Most important, Bayesian nonparametrics provides more flexibility than classical nonparametrics and a more robust analysis than both classical and Bayesian parametric inference.

It deserves to be an important part of the Bayesian paradigm.