Elman network voting system for cyclic system

Abstract

It is important to improve voting system in current software fault tolerance research. In this paper, we propose an Elman network voting system. This is an application of Elman network (a form of recurrent neural network). In time sequential environment, Elman network can predict next state by referencing previous state. Thus, Elman network is especially suitable for cyclic system. Majority voting system is a classical voting system with inherent safety mechanism. Combination of majority voting system and Elman network earns predictive and secure characteristics. Experiment shows that Elman network voting system can give an appropriate advice for disagreement situation after training; this approach performs quite well in small and big turbulent.