Nonlinear Dynamics Analysis of Complex-Valued Neural Networks and its Engineering Applications

点击翻译按钮获取中文摘要

Complex-valued neural networks have been studied from nonlinear dynamics viewpoint to develop the theory and the algorithm for addressing complex-valued information such as wave phenomena and gray-level images. I have proposed a new activation function with continuous nonlinearity for a complex-valued neuron and applied it to gray-level image restoration and graph coloring problem. Numerical results have shown that the performance of complex-valued neural network can be improved by the proposed continuous activation function compared with the conventional discrete activation function.