| 系数正则化回归模型的最优正则参数 |
| |
| 引用本文： | 冯李哲,盛宝怀.系数正则化回归模型的最优正则参数[J].浙江大学学报(理学版),2012,39(2):146-151. |
| |
| 作者姓名： | 冯李哲 盛宝怀 |
| |
| 作者单位： | 1. 浙江财经学院信息学院,浙江杭州,310018
2. 绍兴文理学院数理信息学院,浙江绍兴,312000 |
| |
| 基金项目： | 国家自然科学基金资助项目(10871226) |
| |
| 摘 要： | 研究了具有最小平方损失且正则项为系数正则化的回归问题的误差分析,分别对样本误差和逼近误差作了估计,获得了关于参数γ的误差界;通过选择合适的参数,使得该误差界最优并且得到学习速率.
|
| 关 键 词： | 回归函数 系数正则化回归 覆盖数 再生核hilbert空间 hoeffding不等式 |
On the optimized parameters of coefficient regularized regressions |
| |
| FENG Li-zhe
,
SHENG Bao-huai.On the optimized parameters of coefficient regularized regressions[J].Journal of Zhejiang University(Sciences Edition),2012,39(2):146-151. |
| |
| Authors: | FENG Li-zhe SHENG Bao-huai |
| |
| Institution: | 1.Department of Information,Zhejiang University of Finance & Economics,Hangzhou 310018,China;2.Department of Mathematics,Shaoxing College of Arts and Science,Shaoxing 312000,Zhejiang Province,China) |
| |
| Abstract: | This paper focuses on the error analysis of coefficient regularized algorithm associated with least-square loss function.By estimating the sample error and the approximation error respectively,a bound with respect to the parameter γ is presented and the bound has a minimizer when choosing an appropriate parameter.Meanwhile,learning rates are derived. |
| |
| Keywords: | regression function coefficient regularized regression covering number reproducing kernel Hilbert space Hoeffding’s inequality |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
| 点击此处可从《浙江大学学报(理学版)》浏览原始摘要信息 |
| 点击此处可从《浙江大学学报(理学版)》下载免费的PDF全文 |
