On global asymptotic stability of recurrent neural networks with time-varying delays. (Q1399723)

The authors study the global asymptotic stability of a recurrence neural network of the form model with time-varying delays \[ x'_i(t)=-d_ih_i(x_i(t))+\sum_{j=1}^{n}T_{ij}^0f_j(x_j(t))+ \sum_{j=1}^{n}T_{ij}^{\tau}g_j(x_j(t-r_j(t)))+I_j, \quad i=1,2,\ldots,n. \] This includes some interesting examples like delayed Hopfield neural networks and others. The main result is based on the notion of nonsingular M-matrix and combines topological degree techniques with the construction of a new Lyapunov functional. The activation functions \(f_i\) and \(g_i\) are required to be Lipschitz continuous, but other usual conditions in the related literature, like differentiability, boundedness or monotonicity, are avoided.