A hierarchical mixture model voting system

It is important to improve voting system in current software fault tolerance research. In this paper, we propose a hierarchical mixture model voting system (HMMVS). This is an application of the hierarchical mixtures of experts (HME) architecture. In HMMVS, individual voting models are used as experts. During the training of HMMVS, an Expectation-Maximizing (EM) algorithm is employed to estimate the parameters for HME architecture. Experiments illustrate that our approach performs quite well after training, and better than single classical voting system. We show that the method can automatically select the most appropriate lower-level model for the data and performances are well in voting procedure.