Empirical Bayes nonparametric kernel density estimation

We propose using empirical Bayes on estimated density values to exploit potential similarities among a set of unknown densities. The strengths are that it allows all types of kernel estimators and does not require specification as to the form of similarity.