| 摘要: |
| 本文研究了基于核方法下的在线变化损失函数的回归算法. 利用迭代和比较原则, 得到了算法的收敛速度, 并将该结果推广到了更一般的输出空间. |
| 关键词: 分位数回归 Pinball损失函数 再生核希尔伯特空间 在线算法 |
| DOI： |
| 分类号:O211.6 |
| 基金项目:Supported by by the Special Fund of Basic Scientific Research of Central Colleges (CZQ13015) and the Teaching Research Fund of South-Central University for Nationalities (JYX13023) |
|
| VARYING QUANTILE REGRESSION WITH ONLINE SCHEME AND UNBOUNDED SAMPLING |
|
WANG Bao-bin1, YIN Hong2
|
|
1.School of Mathematics and Statistics, Central South University for Nationalities, Wuhan 430074, China;2.School of Information, Renmin University of China, Beijing 100872, China
|
| Abstract: |
| We consider a kernel-based online quantile regression algorithm associated with a sequence of insensitive pinball loss functions. By iteration method and comparison theorem, we obtain the error bound based on the more general output space. |
| Key words: quantile regression Pinball loss reproducing kernel Hilbert space online algo-rithm |
