《计算机应用》唯一官方网站 ›› 2021, Vol. 41 ›› Issue (12): 3462-3467.DOI: 10.11772/j.issn.1001-9081.2021060998

• 第十八届中国机器学习会议(CCML 2021) • 上一篇    

基于神经正切核的多核学习方法

王梅1,2, 许传海1, 刘勇3,4()   

  1. 1.东北石油大学 计算机与信息技术学院,黑龙江 大庆 163318
    2.黑龙江省石油大数据与智能分析重点实验室(东北石油大学),黑龙江 大庆 163318
    3.中国人民大学 高瓴人工智能学院,北京 100872
    4.大数据管理与分析方法研究北京市重点实验室(中国人民大学),北京 100872
  • 收稿日期:2021-05-12 修回日期:2021-06-29 接受日期:2021-07-05 发布日期:2021-12-28 出版日期:2021-12-10
  • 通讯作者: 刘勇
  • 作者简介:王梅(1976—),女,河北保定人,教授,博士,CCF会员,主要研究方向:机器学习、核方法、模型选择
    许传海(1998—),男,黑龙江鸡西人,硕士研究生,CCF会员,主要研究方向:深度核学习;
  • 基金资助:
    国家自然科学基金面上项目(51774090);黑龙江省博士后科研启动金资助项目(LBH-Q20080);黑龙江省自然科学基金资助项目(LH2020F003);黑龙江省高等教育教学改革重点委托项目(SJGZ20190011)

Multi-kernel learning method based on neural tangent kernel

Mei WANG1,2, Chuanhai XU1, Yong LIU3,4()   

  1. 1.School of Computer and Information Technology,Northeast Petroleum University,Daqing Heilongjiang 163318,China
    2.Heilongjiang Key Laboratory of Petroleum Big Data and Intelligent Analysis (Northeast Petroleum University),Daqing Heilongjiang 163318,China
    3.Gaoling School of Artificial Intelligence,Renmin University of China,Beijing 100872,China
    4.Beijing Key Laboratory of Big Data Management and Analysis Methods (Renmin University of China),Beijing 100872,China
  • Received:2021-05-12 Revised:2021-06-29 Accepted:2021-07-05 Online:2021-12-28 Published:2021-12-10
  • Contact: Yong LIU
  • About author:WANG Mei, born in 1976, Ph. D., professor. Her research interests include machine learning, kernel method, model selection.
    XU Chuanhai, born in 1998, M. S. candidate. His interests include deep kernel learning.
  • Supported by:
    the Surface Program of National Natural Science Foundation of China(51774090);the Postdoctoral Research Startup Foundation of Heilongjiang Province(LBH-Q20080);the Natural Science Foundation of Heilongjiang Province(LH2020F003);the Higher Education Teaching Reform Key Entrusted Project of Heilongjiang Province(SJGZ20190011)

摘要:

多核学习方法是一类重要的核学习方法,但大多数多核学习方法存在如下问题:多核学习方法中的基核函数大多选择传统的具有浅层结构的核函数,在处理数据规模大且分布不平坦的问题时表示能力较弱;现有的多核学习方法的泛化误差收敛率大多为O1/n,收敛速度较慢。为此,提出了一种基于神经正切核(NTK)的多核学习方法。首先,将具有深层次结构的NTK作为多核学习方法的基核函数,从而增强多核学习方法的表示能力。然后,根据主特征值比例度量证明了一种收敛速率可达O1/n的泛化误差界;在此基础上,结合核对齐度量设计了一种全新的多核学习算法。最后,在多个数据集上进行了实验,实验结果表明,相比Adaboost和K近邻(KNN)等分类算法,新提出的多核学习算法具有更高的准确率和更好的表示能力,也验证了所提方法的可行性与有效性。

关键词: 机器学习, 多核学习, 神经正切核, 核对齐, 主特征值比例

Abstract:

Multi-kernel learning method is an important type of kernel learning method, but most of multi-kernel learning methods have the following problems: most of the basis kernel functions in multi-kernel learning methods are traditional kernel functions with shallow structure, which have weak representation ability when dealing with the problems of large data scale and uneven distribution; the generalization error convergence rates of the existing multi-kernel learning methods are mostly O1/n, and the convergence speeds are slow. Therefore, a multi-kernel learning method based on Neural Tangent Kernel (NTK) was proposed. Firstly, the NTK with deep structure was used as the basis kernel function of the multi-kernel learning method, so as to enhance the representation ability of the multi-kernel learning method. Then, a generalization error bound with a convergence rate of O1/n was proved based on the measure of principal eigenvalue ratio. On this basis, a new multi-kernel learning algorithm was designed in combination with the kernel alignment measure. Finally, experiments were carried out on several datasets. Experimental results show that compared with classification algorithms such as Adaboost and K-Nearest Neighbor (KNN), the newly proposed multi-kernel learning algorithm has higher accuracy and better representation ability, which also verifies the feasibility and effectiveness of the proposed method.

Key words: machine learning, multi-kernel learning, Neural Tangent Kernel (NTK), kernel-target alignment, principal eigenvalue ratio

中图分类号: