摘要
针对多源聚合下同时对齐域不变特征较困难而造成分类精度不高的问题,提出基于自监督任务的多源无监督域适应法.该方法引入旋转、水平翻转和位置预测3个自监督辅助任务,通过伪标签性、语义信息的一致性对无标签数据进行自适应的对齐优化.构建新的优化损失函数,减少多域公共类别的分类差异.针对类别不均衡的问题,基于少样本大权重的原则,定义动态权重参数,提高模型的分类性能.在公开的Office-31、Office-Caltech102种基准数据集上,与现有的主流方法进行实验对比.实验结果表明,在类别均衡、不均衡2种情况下,分类精度最高可以提高6.8%.
A multi-source unsupervised domain adaptation method based on self-supervised tasks was proposed aiming at the problem of low classification accuracy due to the difficulty of simultaneously aligning domaininvariant features under multi-source aggregation.The method introduced three self-supervised auxiliary tasks of rotation,horizontal flip and position prediction,and performed adaption alignment optimization on unlabeled data through pseudo-labeling and consistency of semantic information.A new optimized loss was built,and the classification variance of multi-domain common classes was reduced.Dynamic weight parameters were defined to improve the classification performance of the model based on the principle of few samples and large weights for the problem of class-imbalance.Experiments were compared with the existing mainstream methods on the two benchmark data sets,Office-31 and Office-Caltech10.The experimental results show that the classification accuracy can be improved by up to 6.8%in the two cases of class balance and imbalance.
作者
吴兰
王涵
李斌全
李崇阳
孔凡士
WU Lan;WANG Han;LI Bin-quan;LI Chong-yang;KONG Fan-shi(School of Electrical Engineering,Henan University of Technology,Zhengzhou 450001,China;Zhengzhou Railway Vocational and Technical College,Zhengzhou 450001,China)
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2022年第4期754-763,共10页
Journal of Zhejiang University：Engineering Science
基金
国家自然科学基金资助项目(61973103)
河南省优秀青年科学基金资助项目
郑州市协同创新专项资助项目(21ZZXTCX01)。
关键词
自监督任务
类别不均衡
语义信息
权重
域自适应
self-supervised task
class-imbalance
semantic information
weight
domain adaptation
作者简介
吴兰(1981—),女,教授,博士,从事深度学习的研究.orcid.org/0000-0002-2497-6556.E-mail:wulan@haut.edu.cn。
