We focus on semi-supervised classifier, where a decision rule is to be learned from labeled and unlabeled data. A model to semi-supervised classification is proposed to overcome the problem induced by mislabeled sampl...We focus on semi-supervised classifier, where a decision rule is to be learned from labeled and unlabeled data. A model to semi-supervised classification is proposed to overcome the problem induced by mislabeled samples. A new energy function based on robust error function is used in Markov Random Field. Also two algorithms based on iterative condition mode and markov chain monte carlo respectively are designed to infer the label of both labeled and unlabeled samples. Our experiments demonstrate that the proposed method is efficient for real-world dataset.展开更多
基金Acknowledgement: This paper is supported by the National Natural Science Foundation of China (No. 60673190).
摘要We focus on semi-supervised classifier, where a decision rule is to be learned from labeled and unlabeled data. A model to semi-supervised classification is proposed to overcome the problem induced by mislabeled samples. A new energy function based on robust error function is used in Markov Random Field. Also two algorithms based on iterative condition mode and markov chain monte carlo respectively are designed to infer the label of both labeled and unlabeled samples. Our experiments demonstrate that the proposed method is efficient for real-world dataset.