Learning Bayesian network parameters based on iterative learning control

Learning Bayesian network parameters is the basis for Bayesian network to solve practical problems. However, the current methods of learning Bayesian network parameters have problems of large computation capacity, slow rate and so on, especially, more severe in learning parameter with missing data. Therefore, we propose a learning algorithm based on iterative learning control, describe the principle of iterative learning control. The article present the dynamic system of Bayesian network and corresponding updating law, analyze and demonstrate the updating law´s convergence. The numerical simulations show that iterative learning control overcomes the problem of missing data, accelerates convergence rate, algorithm is efficient and worthy.