I. Introduction
The past several decades have witnessed an ever-increasing research interest devoted to the dynamical analysis issues of artificial neural networks (ANNs), which stems mainly from the remarkable superiorities of the ANNs in terms of self-learning capability, fault tolerance, associative memory, and nonlinearity approximation [4], [20], [22], [32], [43]. As abstractions of the biological neural networks, the ANNs aim to simulate certain intelligence activities of human brain, thereby facilitating the resolution of practical problems. Nowadays, the ANNs have been exploited to characterize various systems or phenomena in practical world and found a plethora of successful applications in numerous domains, which include, but are not limited to, wireless communication [6], pattern recognition [3], winemaking technology [25], disease diagnosis [1], robotics, and control [8], [18].