摘要
模糊最小二乘孪生支持向量机模型融合了模糊函数和最小二乘孪生支持向量机算法特性,以解决训练数据集存在孤立点噪声和运算效率低下问题。针对回归过程基于统计学习结构风险最小化原则,对该模型进行L_2范数正则化改进。考虑到大规模数据集的训练效率问题,对原始模型进行了L_1范数正则化改进。基于增量学习特性,对数据集训练过程进行增量选择迭加以加快训练速度。在UCI数据集上验证了相关改进算法的优越性。
Fuzzy least squares twin support vector machines model combined the characteristics of fuzzy function and least squares twin support vector machines algorithm to solve the problem of isolated point noise and inefficiency of training data set.According to regression progress based on the principle of minimizing the risk of statistical learning structure,a regularization of L 2 norm was improved for this model.Taking into account the training efficiency of large-scale data sets,L 1 model regularization was improved on the original model.Finally,based on the incremental learning characteristics,the data set training process was incrementally selected and superposed to speed up training.The superiority of the improved algorithm was verified on the UCI dataset.
作者
唐辉军
杨志民
Tang Huijun;Yang Zhimin(College of Information Engineering,Ningbo Dahongying University,Ningbo 315175,Zhejiang,China;College of Zhijiang,Zhejiang University of Technology,Hangzhou 310024,Zhejiang,China)
出处
《计算机应用与软件》
北大核心
2018年第4期281-286,共6页
Computer Applications and Software
基金
国家自然科学基金项目(10926198)
浙江省公益技术应用研究计划项目(2016C33249)
宁波市自然科学基金项目( 2015A610135
2017A610122)
关键词
最小二乘孪生支持向量机
模糊隶属度
正则化
增量学习
Least squares twin support vector machines
Fuzzy membership
Regularization
Incremental learning
