Modeling, verification and comparison of Zhang Neural Net and gradient neural net for online solution of time-varying linear matrix equation

Abstract

A new recurrent neural network (or say, net), i.e., Zhang Neural Network (ZNN), is recently proposed by Zhang et al for online time-varying matrix equations solving. Theoretical analysis, blocks modeling and verification results of Zhang neural network are investigated in this paper, in addition to the neuralsolver design method and its comparable gradient neural network (GNN). Towards the final purpose of hardware realization, this paper highlights the model building and convergence illustration of ZNN model in comparison with GNN. The verification results substantiate the feasibility and efficacy of ZNN model for online time-varying linear matrix equations solving. Index Terms-time-varying linear matrix equations, recurrent neural networks (RNN), gradient-based neural networks, blocks modeling, verification and comparison.