论文部分内容阅读
该文基于全反馈高阶关联神经网络优化理论,提出了一种将神经网络优化方法应用于ARMA谱估计(ARMA-NNO法)的理论框架.该方法与迄今为止所见方法的区别在于,它直接面对ARMA扩展的Yule?Walker方程的非线性,同时估计出模型的AR和MA两部分参数.描述估计质量的加权均方误差被当作神经网络能量函数,从而导出了ARMA-NNO法的Lyapunov方程.文中讨论了此法的实现方案,给出了几个谱估计实例,通过与其它几种ARMA谱估计方法的比较,证明了它的有效性.“,”A cove approach for ARMA spectral estimation based on Neural Network Optimization (NNO) technique is proposed in the present paper. To demonstrate the feasibility of the NNO approach, a High-Order Interconnected Neural Network model is chosen and the relation between the MSE criterion and the Lyapunov Function is also established The implemenentation of the approach is described together with some guidelines. A few ARMA spectral estimation examoples are given and the advantages of the NNO approach over conventional ones are illustrated.