Fast algorithm for diagonal recurrent neural networks control system

Convergence Theorem 1 in Ref. was given for three layers diagonal recurrent neural networks (DRNN) by introducing a Lyapunov function. Because the essential condition to Theorem 1 was neglected upper limits of learning rates for every weight vectors and matrix were not attained. Much bigger learning rates of all weight vectors and matrix are deduced precisely on the basis of convergence theorem 1 in Ref. , so a fast iterative algorithm is obtained. Simulation results are included.