复交替投影神经网络

首先对ＭarksII等人提出的交替投影神经网络进行了研究，将其应用范围从实数域拓广到复数域，并证明了其稳态收敛性，接着研究了复数域的中替投影神经网络（不妨称此情形下的网络为复交替投影响神经网络），得到了几个有用结论，并从数学上对这些结构进行了严格证明，最后，为了验证文中物理分析的正确性，设计了一个仿真实验，仿真实验结果表明文中的理论与验结果完全吻合，实际上，复交替投影神秒仅可用于信号处理中的带限信号外推，还可用于选频，陷波等场合。

Complex Alternating Projection Neural Networks

After studying alternating projection neural networks proposed byMarks I etc.. This paper extends it from the field of real unmbers to the field of complex numbers, gives its convergent condition and the convergence proof of APNN. Then several useful conclusions are obtained by studying alternation projection neural networks operating in the field of complex numbers (the extended network is called complex alternating projection neural networks), and the strict mathematical proofs are provided for these conclusions. Finally, a simulational experiment is designed to validate the theoretical analysis, the experimental results demonstrate a reasonably good agreement between theory and experiment. In practice complex alternating projection neural networks can be widely applied to signal processing, such as band-limited signal extrapolation, frequency selector and trap filter.