Fast and Exact Leave-One-Out Analysis of Large-Margin Classifiers

Fast and Exact Leave-One-Out Analysis of Large-Margin Classifiers
复制标题

大余量分类器的快速准确留一分析

DOI:
10.1080/00401706.2021.1967199
复制
发表时间:
2022
期刊:
影响因子:
2.5
通讯作者:
Zou, Hui
Zou, Hui
中科院分区:
工程技术3区
文献类型:
--
作者:
Wang, Boxiang;Zou, Hui

文献摘要

参考文献

被引文献

相似文献

受Golub-Heath-Wahba岭回归公式的启发,我们首先提出了一个新的留一引理的核支持向量机(SVM)和相关的大间隔分类器。然后,我们使用引理设计一种新颖而有效的算法,名为“magicsvm”,用于训练核SVM和相关的大间隔分类器,并计算精确的留一交叉验证误差。通过“magicsvm”,留一法分析的计算成本与在训练数据上拟合单个SVM的计算成本是相同的数量级。我们表明,“magicsvm”是远远快于国家的最先进的SVM求解器的基础上广泛的模拟和基准测试的例子。同样的思想也被用来提高核分类器的V折交叉验证的计算速度。
Motivated by the Golub–Heath–Wahba formula for ridge regression, we first present a new leave-one-out lemma for the kernel support vector machines (SVM) and related large-margin classifiers. We then use the lemma to design a novel and efficient algorithm, named “magicsvm,” for training the kernel SVM and related large-margin classifiers and computing the exact leave-one-out cross-validation error. By “magicsvm,” the computational cost of leave-one-out analysis is of the same order of fitting a single SVM on the training data. We show that “magicsvm” is much faster than the state-of-the-art SVM solvers based on extensive simulations and benchmark examples. The same idea is also used to boost the computation speed of theV-fold cross-validation of the kernel classifiers.
DOI: 10.1080/03610927508827233
发表时间: 1975
影响因子: 0.8
作者:
G. Wahba;S. Wold
通讯作者: S. Wold
DOI: 10.1137/080716542
发表时间: 2009-01-01
影响因子: 2.1
作者:
Beck, Amir;Teboulle, Marc
通讯作者: Teboulle, Marc
DOI: 10.1080/01621459.2018.1424632
发表时间: 2020-01-02
影响因子: 3.7
作者:
Rosset, Saharon;Tibshirani, Ryan J.
通讯作者: Tibshirani, Ryan J.
DOI: 10.1214/20-aos1965
发表时间: 2021-02-01
影响因子: 4.5
作者:
Barber, Rina Foygel;Candes, Emmanuel J.;Tibshirani, Ryan J.
通讯作者: Tibshirani, Ryan J.