Finite-time convergence analysis and verification of improved ZNN for real-time matrix inversion

An improved Zhang neural network (IZNN) together with its implementation architecture is presented and investigated for matrix inversion in real time. It is theoretically proved that the IZNN model converges to the theoretical inverse within finite time. In addition, the upper bound of the convergence time is derived analytically via Lyapunov theory. Compared with conventional gradient-based neural network (GNN), the IZNN model depicted in implicit dynamics can coincide better with practical systems. Besides, the IZNN model can achieve excellent finite-time convergence, as compared to the existing recurrent neural networks (RNN), specially the GNN model and the original ZNN (OZNN) model. Simulative results further verify the efficacy and superiority of the IZNN model.