基于自编码器的零样本学习方法研究进展  

Research Progress of Zero-Shot Learning Method Based on AutoEncoder

作　　者：杨晨曦 左劼[1] 孙频捷[2] YANG Chen-xi;ZUO Jie;SUN Pin-jie(College of Computer Science,Sichuan University,Chengdu 610065;Shanghai University of Political Science and Law,Shanghai 200000)

机构地区：[1]四川大学计算机学院,成都610065 [2]上海政法学院,上海200000

出　　处：《现代计算机》2020年第1期48-52,共5页Modern Computer

基　　金：国家重点研发计划项目

摘　　要：零样本学习不需要额外的样本标注以及训练数据,便能有效地扩展模型识别能力,这个特点使之成为一个热门而实用的研究领域。近年来,基于自编码器的零样本学习方法不仅能有效解决映射域偏移问题,更是在分类、识别等任务上带来精度的显著提升,这引起学者们的广泛关注。通过对自编码器应用到零样本学习上的方式进行分析归纳,为基于自编码器的零样本学习方法的进一步研究提供理论基础。Zero-Shot Learning can effectively expand the model recognition ability without additional sample annotation and training data,which makes it a hot and practical research field.In recent years,the Zero-Shot Learning method based on AutoEncoder can not only effectively solve the problem of projection domain shift,but also bring significant improvement in the accuracy of classification,recognition and other tasks,which has attracted extensive attention of scholars.Analyzes and sums up the ways of AutoEncoder applied to Zero-Shot Learning,which provides a theoretical basis for further research on Zero-Shot Learning method based on AutoEncoder.

关 键 词：零样本学习 自编码器 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]