摘要
极限学习机ELM(Extreme learning machine)作为一个有竞争力的机器学习算法,以其简单的理论和易于实施的特点吸引了越来越多学者的关注。近来,针对噪音及离群数据,研究人员提出了相关的研究算法,然而如何将ELM更好地应用在含有噪音及离群数据的分类问题中仍是一个重要的研究课题。基于数据的信息关联的技术思想提出一种修正的模糊极限学习机(MFELM)。MFELM的优势在于:1)MFELM在处理噪音及离群数据的分类问题时能够保持ELM处理正常数据分类问题的良好性能;2)适用于ELM的激活函数或核函数同样适用于MFELM模型;3)根据不同的需求给每个数据样本分配不同的隶属度,MFELM可以推广到代价敏感学习中。通过使用UCI数据集和普遍应用的人脸数据集进行实验,实验结果表明该提出的算法显著提高了ELM的分类能力并优于其他算法。
As a competitive machine learning algorithm,Extreme Learning Machine( ELM) attracts more and more scholars' attention with its simple theory and easy implementation. Recently,researchers on noise and outlier data have proposed relevant research algorithms. However,how to use ELM better in classification problems with noise and outlier data is still an important research topic. This paper proposes a modified fuzzy extreme learning machine( MFELM)based on the technical idea of information related. The advantages of MFELM are as follows: MFELM can maintain the good performance of ELM processing normal data classification when dealing with the classification of noise and outlier data; the activation or kernel functions for the ELM also apply to the MFELM model; MFELM can be extended to costsensitive learning by assigning different memberships to each data sample according to different requirements.Experiments on UCI datasets and universally applied face datasets show that the proposed algorithm improves the classification ability of ELM significantly and is superior to other algorithms.
出处
《计算机应用与软件》
2017年第5期234-240,共7页
Computer Applications and Software
基金
国家自然科学基金项目(61373127)
关键词
极限学习机
不平衡数据
信息关联
特征映射
Extreme learning machine Imbalanced data Information related Feature mapping
