论文部分内容阅读
大多统计模型的输出与输入都是高度非线性和线性相叠加的关系,为了更好地实现数据驱动的研究,本文提出了一种隐含层组合型的ELM(Extreme Learning Machine with Hybrid Hidden Layer,HHL-ELM)神经网络。该HHL-ELM神经网络在传统的ELM网络的隐含层中增加一个特殊的节点,该特殊节点的激活函数与隐含层其他节点激活函数不同,从而形成了一种隐含层组合的网络结构,试图增强ELM网络模型的输出。同时,本文利用UCI标准数据集中的Housing数据集进行了测试,并通过工业应用实例进行了验证。最后进行了模型对比,结果表明HHL-ELM网络在处理复杂数据时具有精度高的特点,为神经网络发展及其应用提供了新思路。
Most of the statistical model output and input are highly nonlinear and linear superposition relationship, in order to better achieve the data-driven research, this paper presents a hidden layer combination ELM (Extreme Learning Machine with Hybrid Hidden Layer, HHL-ELM) neural network. The HHL-ELM neural network adds a special node in the hidden layer of the traditional ELM network. The activation function of this special node is different from the activation functions of other nodes in the hidden layer, so that a network structure with a hidden layer combination is formed , Trying to enhance the ELM network model output. At the same time, this paper tests the Housing dataset in UCI standard dataset and verifies it through industrial application examples. Finally, the model is compared. The results show that the HHL-ELM network has the characteristics of high precision when dealing with complex data and provides a new idea for the development of neural network and its application.