摘要
Let （X1, θ1）,…, （Xn, θn） be iid. samples, usually called training samples from the population （X,θ） which is an Rd×（1, 2, …, s）-valued random vector, and let ρ be a distance function equivalent to the Euclidean one in Rd. For X = x, rank the （Xj, θj）, j = 1,2,…,n, according to increasing values of ρ（Xj, x） (ties are broken
