U-processes: Rates of convergence (Q578728)

Let \(\xi_ 1,\xi_ 2,..\). be independent, identically distributed random variables and denote by \[ S_ n(f)=\sum_{1\leq i\neq j\leq n}f(\xi_ i,\xi_ j) \] the U-statistic with respect to the kernel f. The authors obtain almost sure convergence results for \(S_ n(f)\) uniformly over \(f\in F\) where F belongs to certain classes of kernels. Assumptions and proofs are motivated by the corresponding theory for empirical processes, though there are several significant differences in this case. Finally, an application to cross validation in density estimation is given.