A noise-tolerant Z-type neural network for time-dependent pseudoinverse matrices

Differing from the well-known gradient neural network (GNN) and the conventional Z-type neural network (CZNN), in this paper, we first put forward and generalize the improved Z-type neural network, termed noise-tolerant Z-type neural network (NTZNN), to compute the real-time-dependent matrix pseudoinverse under noisy environments. Theoretical analyses substantiate that the presented NTZNN has the capability of globally exponential convergence, and can resist various noises simultaneously. For comparative purposes, the gradient neural network and the conventional Z-type neural network are also presented and analyzed to handle the same time-dependent problem. Finally, numerical examples and results further substantiate the superior performance of the proposed NTZNN for computing the real-time-dependent matrix pseudoinverse in the situation of various types of noise, as comparing with the GNN and CZNN.