摘要
假若一个以点积为自变量的多项式不是再生核,则它无法在机器学习的核方法中使用.解决此问题的办法之一是匹配另外一个点积的多项式,使两者乘积成为再生核.在一定条件下,通过解一系列的不等式,得到匹配多项式存在的充分必要条件,并探讨与此条件相关的数列和生成函数列的性质.
If a polynomial with the dot product as its independent variable is not a reproducing kernel, there will be no way to use it in kernel method for machine learning. One of the dealing methods is matching it with another dot-product polynomial such that their product is areproducing kernel. Under certain conditions and by means of solving a sequence of inequality systems, a sufficient and necessary condition of the existence of the matching polynomials is found and the sequence relevant to this condition and the properties of the generated function sequence are explored.
作者
刘建强
LIU Jian-qiang(School of Mathematics and Statistics,Ningxia University,Yinchuan750021,China)
出处
《兰州理工大学学报》
CAS
北大核心
2019年第2期149-154,共6页
Journal of Lanzhou University of Technology
基金
宁夏高校科学研究项目(NCY2018047)
关键词
点积核
再生核
多任务核
连分式
函数列
dot product kernel
reproducing kernel
multitask kernel
continued fraction
function sequence