Multi-thread search with deterministic annealing EM algorithm

The EM algorithm is an efficient algorithm to obtain the ML estimate for incomplete data, but has the local optimality problem. The deterministic annealing expectation-maximization (DAEM) algorithm was once proposed to solve the problem, but is not guaranteed to obtain the global optimum since it employs a single token search. The paper investigates the possibility of the multiple-thread search with the DAEM algorithm for a Gaussian mixture. The experiments showed the minimal beam size to guarantee the global optimality is not so large for a Gaussian mixture, and the solution quality of the beam DAEM algorithm always exceeds the EM and DAEM algorithms.