0


在线学习

Online Learning
课程网址: http://videolectures.net/bootcamp2010_paredes_ol/  
主讲教师: Roberto Paredes
开课单位: 瓦伦西亚技术大学
开课时间: 2010-08-05
课程语种: 英语
中文简介:
这篇演讲的目的是展示对在线学习方法的第一次探讨。它开始提出线性感知器算法和一些推导。内核感知器算法也受到激励。针对二元分类、回归和多类问题,提出了一种被动攻击性在线学习算法。本文提出并讨论了PA的线性和核版本,特别注意更新规则的推导。一个重要的问题是能够控制模型的复杂度,为此提出了不同的在线预算算法,首先是传统的感知器和推导,其次是PA算法。最后给出了一些在线学习应用的例子,以激励读者在不同的真实场景中应用这些技术。
课程简介: This talk aim at presenting a first approaching to the Online Learning methodology. It starts presenting the linear Perceptron algorithm and several derivations. The kernel Perceptron algorithm is also motivated. The Passive- Aggressive (PA) Online Learning algorithm is then presented for binary classification, regression and multi-class problems as well. Linear and kernel version of PA are presented and discussed playing an especial attention to the update rules derivation. An important issue is to be able to control the model complexity, to this end different budgeted online algorithms are proposed, first for the conventional Perceptron and derivations and second, for the PA algorithm. Finally some examples of online learning applications are presented in order to motivate the audience to the application of these techniques in different real scenarios.
关 键 词: 在线学习; 二进制; 感知器
课程来源: 视频讲座网
最后编审: 2020-04-08:cjy
阅读次数: 49