Minka s Bayes Point Machine. Applying Expectation Propagation to inference.

infinite HMM Refer to Beal s Paper Applying Gibbs Sampling in inference the number of states of HMM is changeable in this algorithm, based on Dirichlet Proce