Stability analysis of a class of discrete-time recurrent neural networks: An LMI approach

Stability analysis of discrete-time recurrent neural networks is seldom researched at present. By using the state space extension method, discrete-time recurrent neural networks with sector-type monotone nonlinear activation functions, also known as recurrent multilayer perceptrons (RMLPs), were converted to the forms represented as linear differential inclusions (LDIs). Stability conditions of LDIs were transformed into some linear matrix inequalities (LMIs) which were solved by MATLAB/LMI TOOLBOX to determine whether RMLPs were Lyapunov stable or not. The proposed approach can also be applied to other forms of recurrent neural networks (RNNs).