摘要
针对原始极限学习机不适于解决在线预测的回归问题,提出一种样本增量型极限学习机算法。极限学习机是一种新型的单隐层前向神经网络,随机设置输入权值和隐层阈值,解析计算获得输出权值。所提算法以极限学习机为基础,引入样本增量的思想,即根据样本之间的实际增量来调整网络的输入权值和阈值,使极限学习机具有样本自适应性和在线辨识能力。通过在UCI数据集和锅炉NOx排放浓度建模上的应用,实验结果表明算法具有良好的回归能力和泛化能力,并且对于解决在线建模问题是有效的。
Because the original extreme learning machine is not suitable to solve online-predictive regression problem, a kind of sample increment extreme learning machine(SI-ELM) is proposed in this paper.Extreme learning machine(ELM) is a novel single hidden layer forward neural network.Its input weights and hidden layer bias are generated randomly and the output weights are analytically determined based on least square method.Based on the extreme learning machine, the proposed algorithm introduces the idea of sample increment to adjust the input weights and hidden layer biases, which makes the ELM owning the adaptivity and online identification ability for input sample data.The proposed algorithm is applied to solve benchmark problems of UCI datasets and build the NOx emission model of circulation fluidized bed boiler.The experiment results show that the SI-ELM has good regression ability and generalization ability.It is an effective online sequential machine learning tool for online model problems.
作者
马云鹏
王贺琦
唐浩桁
MA Yunpeng;WANG Heqi;TANG Haohang(School of Information Engineering,Tianjin University of Commerce,Tianjin 300134,China)
出处
《自动化与仪器仪表》
2021年第7期5-8,12,共5页
Automation & Instrumentation
基金
国家自然科学基金(No.61573306)
天津市自然科学基金(No.20JCQNJC00430)
天津市创新训练项目(No.202010069066)。
关键词
极限学习机
样本增量
循环流化床锅炉
泛化能力
在线模型
extreme learning machine
sample incremental
circulation fluidized bed boiler
generalization ability
online model
