An Anti-Disturbance Integral Recursive Neural Network for Solving Time-Varying Matrix Inversion

To solve the time-varying matrix for inverse problem, an anti-disturbance integral recursive neural network (ADIRNN) is propounded in this study. The AD-IRNN method with different activation functions is exploited and analyzed, which proves that the proposed method is global convergent. Furthermore, because of the integral-type design formula, the ADIRNN method has strong robustness and can resist disturbances effectively. By incorporating an integral term, the norm-based error function of the AD-IRNN method can converge to zero, and the convergence rate is faster than that of the zeroing neural network. Comparative experiments verify that the AD-IRNN method has the superiority of solving the time-varying matrix for inverse problem.