α混合样本下多核正则分类器的学习速率  

Learning rates for multi-kernel regularized classifiers with α mixing observations

作　　者：邢星[1] 曹飞龙[1] 

机构地区：[1]中国计量学院理学院,浙江杭州310018

出　　处：《中国计量学院学报》2010年第4期349-352,共4页Journal of China Jiliang University

基　　金：国家自然科学基金资助项目(No.60873206)

摘　　要：给出了Tikhonov正则算法的多核正则分类器在α混合样本下的错分误差的界.利用α混合样本的Bernstein不等式,得到了α混合样本情况下的两个相关结果,然后将这两个结果应用到误差分解,并得到学习速率的估计.Tikhonov regularization scheme is one of the most common classification algorithms.Our main purpose was to estimate the bound of excess misclassification error of multi-kernel classfiers with α mixing observations based on the algorithms.Using Bernstein inequality for α mixing sequences,two related results with α mixing observarions were obtained.Furthermore,the two results were applied to error decomposition and the learning rates are implied.

关 键 词：分类算法 多核正则分类 学习速率 α混合 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]