基于改进梯形网络的半监督虚拟对抗训练模型  

Semi-supervised virtual adversarial training model based on improved ladder network

作　　者：莫建文[1] 贾鹏 MO Jianwen;JIA Peng(School of Information and Communication Engineering,Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学信息与通信学院,广西桂林541004

出　　处：《桂林电子科技大学学报》2020年第4期321-327,共7页Journal of Guilin University of Electronic Technology

基　　金：国家自然科学基金(61661017,61967005,U1501252);桂林电子科技大学研究生教育创新计划(2019YCXS020)。

摘　　要：为了提高半监督深层生成模型的分类精度,提出一种基于改进梯形网络的半监督虚拟对抗训练模型。该模型在梯形网络框架的基础上,以mix_up数据增强和虚拟对抗训练相结合的模式训练分类器。用mix_up对训练数据做增强处理得到新的扩展数据,以解决半监督分类模型有标记样本较少的问题,并且对梯形网络框架施加虚拟对抗噪声,通过构建平滑性正则化约束,提高模型的泛化能力。模型以有标记数据的分类损失、未标记数据的重构损失和虚拟对抗损失相结合的方式调整参数,训练得到分类器。模型分别在MNIST数据库、SVHN数据库上进行实验,并且与其他半监督深层生成模型进行对比,结果表明,该模型能增强泛化能力,提高分类精度。In order to enhance the generalization ability and improve the classification performance accuracy of the semi-supervised deep generation model,a semi-supervised virtual adversarial training model based on improved ladder network is proposed.The model trains the classifier with a combination of mix_up data augmentation and virtual adversarial training based on the ladder network.The model use mix_up to enhance the training data to obtain new extended data to solve the problem of fewer labeled samples in the semi-supervised classification model,and the virtual adversarial noise is applied to the ladder network,and the generalization ability of the model is improved by constructing smooth regularization constraints.The model adjusts the parameters by combining the classification loss of labeled data,the reconstruction loss of unlabeled data and the virtual adversarial loss,and trains to obtain a classifier.The model is tested on MNIST and SVHN respectively.In comparison with other semi-supervised deep generation models,test results show that the model can effectively enhance generalization ability and improve the accuracy of image classification.

关 键 词：半监督分类 梯形网络 数据增强 虚拟对抗训练 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]