Efficient basis selection for smoothing splines via rotated lattices

Author:

Diao Huaimin1ORCID,Ai Mengtong2,Tian Yubin1,Yu Jun1ORCID

Affiliation:

1. School of Mathematics and Statistics Beijing Institute of Technology Beijing China

2. Department of Statistics University of Michigan Ann Arbor, Michigan USA

Abstract

Fitting a smoothing spline model on a large‐scale dataset is daunting due to the high computational cost. In this study, we develop an efficient basis selection method for smoothing spline calculation. The key idea is to force a nonparametric function in an infinite‐dimensional functional space to reside in a relatively small and finite‐dimensional model space without the loss of too much prediction accuracy. Such an approximation naturally allows for much faster numerical calculation, especially for large datasets. Among various basis selection methods, space‐filling basis selection has been proven to be more efficient since its model space dimension is smaller than that of others. Despite algorithmic benefits, most of the space‐filling selection methods only take the overall space‐filling property into account. These methods may be less efficient when the underlying response surface is not isomorphic. To overcome this obstacle, we develop an efficient algorithm to improve projective uniformity for space‐filling basis selection. It has been proved that the proposed estimator has the same convergence rate as the full bases estimator. Compared with the standard approach, the proposed method significantly reduces the computational cost. Simulation and real data studies demonstrate the efficiency and superiority of the proposed method.

Funder

National Natural Science Foundation of China

Natural Science Foundation of Beijing Municipality

Beijing Institute of Technology Research Fund Program for Young Scholars

Publisher

Wiley

Subject

Statistics, Probability and Uncertainty,Statistics and Probability

Cited by 1 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

1. Nonparametric Additive Models for Billion Observations;Journal of Computational and Graphical Statistics;2024-03-19

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3