论文部分内容阅读
为了提高太阳黑子预测预报的精度,提出固定型极限学习过程神经网络(FELM-PNN)和增量型极限学习过程神经网络(IELM-PNN)两种学习算法.FELM-PNN的隐层节点数目固定,使用SVD求解隐层输出矩阵的Moore-Penrose广义逆,通过最小二乘法计算隐层输出权值;IELM-PNN逐次增加隐层节点,根据隐层输出矩阵和网络误差计算增加节点的输出权值.通过Henon时间序列预测验证了两种方法的有效性,并实际应用于第24周太阳黑子平滑月均值的中长期预测预报中.实验结果表明,两种方法的预测精度均有一定程度的提高,IELM-PNN的训练收敛性优于FELM-PNN.
In order to improve the accuracy of forecasting of sunspots, two learning algorithms, FELM-PNN and IELM-PNN, are proposed. The number of hidden nodes in FELM-PNN is fixed , The SVD is used to solve the Moore-Penrose generalized inverse of the hidden layer output matrix, and the hidden layer output weights are calculated by the least square method. The IELM-PNN sequentially increases the hidden layer nodes and increases the output weights of nodes according to the hidden layer output matrix and network error The validity of the two methods was verified by Henon time series prediction, and was applied to mid-long term prediction and forecast of smooth monthly average of sunspots at week 24. The experimental results show that the prediction accuracy of both methods are improved to a certain extent , IELM-PNN has better training convergence than FELM-PNN.