论文部分内容阅读
针对基于半监督学习的分类器利用未标记样本训练会引入噪声而使得分类性能下降的情形,文中提出一种具有噪声过滤功能的协同训练半监督主动学习算法.该算法以3个模糊深隐马尔可夫模型进行协同半监督学习,在适当的时候主动引入一些人机交互来补充类别标记,避免判决类别不相同时的拒判和初始时判决一致即认为正确的误判情形.同时加入噪声过滤机制,用以过滤由机器自动标记的可能是噪声的样本.将该算法应用于人脸表情识别.实验结果表明,该算法能有效提高未标记样本的利用率并降低半监督学习而引入的噪声,提高表情识别的准确率.
A semi-supervised learning algorithm based on semi-supervised learning is proposed in this paper, which uses the unlabeled samples to introduce noise into the classifier to reduce the classification performance. A semi-supervised active learning algorithm with noise filtering is proposed in this paper. Kov model collaborative semi-supervised learning, when appropriate, take the initiative to introduce some human-computer interaction to supplement the category marker, to avoid the rejection of the judgment category is not the same as the initial judgment of the same judgment that the correct case of misjudgment. At the same time adding noise filtering Which is used to filter out the possible noise samples automatically tagged by the machine.The algorithm is applied to facial expression recognition.The experimental results show that the algorithm can effectively improve the utilization of unlabeled samples and reduce the noise introduced by semi-supervised learning , Improve the accuracy of expression recognition.