Maximum Likelihood Parameter Estimation for ARMAX Models Based on Stochastic Gradient Algorithm

摘要

In this paper, a modified maximum likelihood based stochastic gradient parameter estimation algorithm is proposed for ARMAX models. First, a maximum likelihood based iterative algorithm is proposed. Since the iterative algorithm has heavy computational efforts, a modified maximum likelihood based stochastic gradient parameter estimation algorithm is developed. This algorithm can estimate the unknown parameters and the unknown noise simultaneously. Compared with the traditional on-line maximum likelihood parameter estimation algorithm, the algorithm in this paper has less computational efforts and more accurate estimation accuracy. The simulation results indicate that the proposed algorithm is effective.