Fitting data with continuous phase-type distributions via deterministic annealing EM algorithm

To overcome the initial parameters sensitive problem in the conventional expectation-maximization(EM) algorithm for phase-type distributions(PH) data fit,a PH distributions data fit method based on the deterministic annealing EM algorithm is proposed.The detailed theoretical inference process of the method is given.The method is compared with standard EM algorithm through two benchmarks.Contrast results show that the proposed method can obtain better estimates free of the initial parameter values and reduce the possibility of getting into local maximum.