Semi-supervised Gaussian random field transduction and induction

作     者：Yangqiu SONG Jianguo LEE Changshui ZHANG Shiming XIANG 

作者机构：State Key Laboratory on Intelligent Technology and SystemsTsinghua National Laboratory for Information Science and Technology(TNList)Department of AutomationTsinghua UniversityBeijing 100084China 

出 版 物：《Frontiers of Electrical and Electronic Engineering in China》 (中国电气与电子工程前沿（英文版）)

年 卷 期：2008年第3卷第1期

页      面：1-9页

学科分类：07[理学] 0701[理学-数学] 070101[理学-基础数学] 

基　　金：This work was supported by the Basic Research Foundation of Tsinghua National Laboratory for Information Science and Technology(TNList) 

主　　题：Gaussian process Gaussian random field semi-supervised learning graph based learning 

摘      要：This paper proposes a semi-supervised inductive algorithm adopting a Gaussian random field(GRF)and Gaussian *** introduce the prior based on graph *** regularization term measures the p-smoothness over the graph.A new conditional probability called the extended Bernoulli model(EBM)is also *** generalizes the logistic regression to the semi-supervised case,and especially,it can naturally represent the *** the training phase,a novel solution is given to the discrete regularization framework defined on the *** the new test data,we present the prediction formulation,and explain how the margin model affects the classification boundary.A hyper-parameter estimation method is also *** results show that our method is competitive with the existing semi-supervised inductive and transductive methods.