论文部分内容阅读
单层感知器神经网络模型是多层感知器神经网络———ＢＰ网络的基础，对单层感知器学习算法的改进是进行ＢＰ网络学习算法改进的基础。把带遗忘因子的递推最小二乘辩识算法的原理应用到单层感知器的学习算法中，提出了单层感知器的改进学习算法。这一改进算法克服了常规学习算法不适于在线学习的缺点。仿真实验的结果证实，基于改进学习算法的单层感知器完全可以满足线性系统在线辨识的要求。最后分析了这种改进算法的优点及其具有这些优点的原因
The single-layer perceptron neural network model is the basis of multi-layer perceptron neural network --- BP network. The improvement of learning algorithm of single-layer perceptrons is the basis of BP network learning algorithm improvement. The principle of recursive least squares identification algorithm with forgetting factor is applied to learning algorithm of single-layer perceptron, and an improved learning algorithm of single-layer perceptron is proposed. This improved algorithm overcomes the shortcomings that conventional learning algorithms are not suitable for online learning. Simulation results show that single-layer sensor based on improved learning algorithm can meet the requirements of on-line identification of linear systems. Finally, the advantages of this improved algorithm and the reasons for these advantages are analyzed