“…The joint compression and classification algorithm developed by Oehler, Gray, Perlmutter, Olshen, et al [13,14,16,17,15,18,19], is referred to as Bayes VQ. The basic assumption is that a training sequence L = {(x i , y i ), i = 1, 2, ..., L} is a realization of a random process {(X i , Y i ), i = 1, 2, ...} with (X i , Y i ) obeying a common but unknown distribution P XY on (X, Y ) ∈ A X × A Y .…”