No abstract

In this paper, we propose a condition on global asymptotic stability for recurrent type complex-valued neural networks with a class of activation functions. Those networks have states, connection weights, and activation functions, which are all complex-valued. Such networks have been studied concerning their abilities of information processing, because of their attractive features which are not existent in real-valued counterparts. In particular, the activation function of a complex-valued neuron is an important factor that characterizes the dynamics of complex-valued neural networks. We propose a class of activation functions which ensure the global asymptotic stability, together with a condition on the connection weights. Furthermore, we apply the proposed dynamical networks and the stability condition to a convex programming problem with nonlinear constraints.

No abstract

No abstract

可， IntroductionAs a psychological viewpoint to investigate positive interests toward the sea in Japanese marine stu ・ dents ， we have conducted