基于Co-Teaching的噪声标签深度学习  

Deep Learning with Noisy Labels Based on Co-Teaching

作　　者：夏强强 李菲菲[1] XIA Qiangqiang;LI Feifei(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《电子科技》2024年第11期1-6,共6页Electronic Science and Technology

基　　金：上海市高校特聘教授(东方学者)岗位计划(ES2015XX)。

摘　　要：大规模数据在人为标记时易出现标记误差,导致数据集存在噪声标签,影响深度神经网络模型的泛化。Co-teaching等现行研究方法中的样本选择机制易使噪声样本流入被选的干净标签样本子集,在训练中难以较好地控制深度神经网络模型对被选干净样本子集的拟合。因此,文中提出一个基于Co-teaching改进的新算法。该方法通过增加两个正则化损失来分别避免模型过于信任某单一类别和陷入局部最优解中。此外,引入大学习率衰减训练方法使模型在训练初期更倾向学习干净标签样本特征以得到较好的模型参数。与Co-teaching结果相比,文中模型在20%和50%对称噪声以及45%非对称噪声环境下,在MNIST、CIFAR-10合成噪声数据集及Animal10N现实数据集上的性能均取得了提升。When large-scale data is labeled artificially,labeling errors are easy to occur,which leads to the existence of noise labels in data sets,and further affects the generalization of deep neural network models.The sample selection mechanism in the existing research methods such as Co-teaching makes the noise samples easy to flow into the selected clean label sample subset,and it is difficult to control the deep neural network model's fitting to the selected clean sample subset in training.Therefore,this study presents a novel algorithm that improves upon Co-teaching.In this method,two regularization losses are added to prevent the model from placing too much trust in a single class and falling into a local optimal solution respectively.Additionally,the introduction of high learning rate attenuation training method makes the model more inclined to learn clean label sample features in the initial training to get better model parameters.Compared with the results of Co-teaching,the performance of the proposed model is improved on MNIST,CIFAR-10 synthetic noise data set and Animal10N realistic data set under 20%and 50%symmetric noise and 45%asymmetric noise environment.

关 键 词：深度学习 卷积神经网络 图像分类 噪声标签数据 标签噪声学习 Co-teaching训练 学习率 鲁棒损失函数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]