This demo nstrates the use of the reversible jump MCMC simulated annealing for neural networks. This algorithm enables us to maximise the joint posterior distribution of the network parameters and the number of basis function. It performs a global se