Existence and stability of periodic solutions for Hopfield neural network equations with periodic input (Q1599964)

Consider \(F\in C^2(\mathbb{R}^n, \mathbb{R})\), \(T= (T_{ij})_{ij}\), \(T_{ij}\in C(\mathbb{R}, \mathbb{R}^n)\), \(1\leq i\leq n\), \(1\leq j\leq n\), \(h= (h_1,\dots, h_n)\in C(\mathbb{R}^n, \mathbb{R}^n)\) given by \[ h(x(t- \tau))= (h_1(x_1(t- \tau)),\dots, h_n(x_n(t- \tau))),\quad t\in\mathbb{R},\quad \tau\in \mathbb{R}, \] \(I= (I_1,\dots, I_n)\in C(\mathbb{R}, \mathbb{R}^n)\), \(T\) and \(I\) being periodic functions of period \(\omega\) and \(x= (x_1,\dots, x_n)\in \mathbb{R}^n\). The authors study existence and stability of \(\omega\)-periodic solutions to the general Hopfield neural network equations with periodic input and delay argument \[ \dot x= -\text{grad }F(x)+ T(t)h(x(t-\tau))+ I(t)\tag{1} \] and prove that equation (1) has at least one periodic solution of period \(\omega\), then studying the uniqueness and stability of \(\omega\)-periodic solutions. If there are constants \(a>0\), \(b\geq 0\), \(c\geq 0\), \(d\geq 0\) and \(kc< a\) such that \[ x^T\text{grad }F(x)\geq a|x|^2- b,\quad |h(x)|\leq c|x|+ d,\quad x\in \mathbb{R}^n,\tag{2} \] then equation (1) has at least one \(\omega\)-periodic solution. Moreover, if there are \(R_0> 0\), \(a> 2^{-1}k\) such that for any \(|x|> R_0\) we have \[ x^T\text{grad }F(x)\geq 2^{-1} k|x|^2+ m|x|+ a|h(x)|^2\tag{3} \] then equation (1) has at least one \(\omega\)-periodic solution. If the conditions (2) or (3) are satisfied and \[ |h_i'(x)|\leq h,\quad 1\leq i\leq n,\quad\widetilde T(t)= hT(t),\quad \widetilde k= \max_{t\in [0,\omega]}\|\widetilde T(t)\|, \] \[ {\partial^2 F\over\partial x^2} (x)\geq kI_n,\quad x\in \mathbb{R}^n,\quad k>\widetilde k, \] then equation (1) has only one \(\omega\)-periodic solution and the \(\omega\)-periodic solution is globally asymptotically stable.