A Bayesian framework for the analysis of radial basis functions (RBF)
is proposed that accommodates uncertainty in the dimension of the mode
l. A distribution is defined over the space of all RBF models of a giv
en basis function, and posterior densities are computed using reversib
le jump Markov chain Monte Carlo samplers (Green, 1995). This alleviat
es the need to select the architecture during the modeling process. Th
e resulting networks are shown to adjust their size to the complexity
of the data.