基于正规方程的L2,1正则核极限学习机
L2,1-regularized kernel extreme learning machine based on normal equation
吴青 1魏瑶 2马甜露 2武江波2
作者信息
- 1. 西安邮电大学自动化学院,陕西西安 710121;西安市先进控制与智能处理重点实验室,陕西西安 710121
- 2. 西安邮电大学自动化学院,陕西西安 710121
- 折叠
摘要
为了降低核极限学习机的时间复杂度,提出一种基于正规方程的L2,1正则核极限学习机.将L2,1范数引入核极限学习机的目标函数中,利用正规方程法求解L2,1正则核极限学习机的最优输出权值,从而避免模型的过拟合问题,同时提高分类性能.实验结果表明,与传统的核极限学习机相比,所提核极限学习机能够有效减少学习过程中的大量矩阵运算,具有更快的学习速度和更高的分类准确率.
Abstract
In order to reduce the time complexity of the kernel extreme learning machine,the L2.1-regularized kernel extreme learning machine based on the normal equation is proposed.The L2,1-norm is introduced into the objective function of the kernel extreme learning machine,and the opti-mal output weights of the L2.1-regularized kernel extreme learning machine are solved by using the normal equation,which effectively avoids the overfitting problem of the model,as well as improves the classification performance.Experiment results indicate that the proposed kernel extreme learn-ing machine can effectively decrease a large number of matrix operations in the learning process,and has faster learning speed as well as higher classification accuracy than the conventional kernel extreme learning machine.
关键词
极限学习机/核函数/L2,1范数/核极限学习机/正规方程Key words
extreme learning machine/kernel function/L2,1-norm/kernel extreme learning ma-chine/normal equation引用本文复制引用
基金项目
国家自然科学基金项目(51875457)
陕西省自然科学基金项目(2022JQ-636)
陕西省重点研发计划项目(2022GY-050)
出版年
2024
NETL
NSTL
万方数据