论文部分内容阅读
为了提高极限学习机对化工过程的高维数据进行建模的能力,提出了一种基于信息熵优化的学习算法。利用互信息方法判断输入变量与输出变量之间的相关性,通过去除部分与输出变量相关性较弱的输入变量来过滤冗余信息,从而达到降维的目的。然后利用熵权法对输入数据进行加权优化,从而降低输入数据中的离散点对极限学习机模型精确度的影响。因此本文提出了一种基于信息熵的ELM算法。该算法以UCI标准数据集进行测试,并以PTA工业系统数据进行实际验证。实验结果表明,与传统ELM算法相比,优化后的学习算法在处理高维数据时具有稳定性强、建模精度高的特点。从而拓展了神经网络技术在化工领域里的应用。
In order to improve the ability of extreme learning machine to model high dimensional data of chemical processes, a learning algorithm based on information entropy optimization is proposed. The mutual information method is used to judge the correlation between input variables and output variables, and the redundant information is filtered by removing some of the input variables that are less relevant to the output variables to achieve the purpose of dimensionality reduction. Then, the input data is weighted and optimized by the method of entropy, so as to reduce the influence of the discrete points in the input data on the accuracy of the limit learning machine model. Therefore, this paper proposes an ELM algorithm based on information entropy. The algorithm is tested with UCI standard dataset and verified with PTA industrial system data. The experimental results show that compared with the traditional ELM algorithm, the optimized learning algorithm has the advantages of strong stability and high modeling accuracy when dealing with high-dimensional data. Thus expanding the neural network technology in the field of chemical applications.