Exponential Bound for Error Probability in NN-Discrimination

摘要

Let (theta,X) be a (1,2,...,s) x R to the d power-valued random vector and (theta sub 1, x sub 1),...,(theta sub n, x sub n) be iid. samples drawn from (theta,X). The so-called discrimination problem is to find a function of X, usually depending upon (theta sub n, x sub n),...,(theta sub, x sub n), which is used to predict the value of theta. One of the most frequently used approaches is the nearest neighbor discrimination rule.