[1]秦牧轩,荆晓远,吴 飞.基于公共空间嵌入的端到端深度零样本学习[J].计算机技术与发展,2018,28(11):44-47.[doi:10.3969/ j. issn.1673-629X.2018.11.010]
　QIN Mu-xuan,JING Xiao-yuan,WU Fei.End-to-end Deep Zero-shot Learning Based on Co-space Embedding[J].,2018,28(11):44-47.[doi:10.3969/ j. issn.1673-629X.2018.11.010]
点击复制

基于公共空间嵌入的端到端深度零样本学习()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
28
期数:
2018年11期
页码:
44-47
栏目:
智能、算法、系统工程
出版日期:
2018-11-10

文章信息/Info

Title:
End-to-end Deep Zero-shot Learning Based on Co-space Embedding
文章编号:
1673-629X(2018)11-0044-04
作者:
秦牧轩; 荆晓远; 吴 飞
南京邮电大学 自动化学院,江苏 南京 210003
Author(s):
QIN Mu-xuan; JING Xiao-yuan; WU Fei
School of Automation,Nanjing University of Posts and Telecommunications,Nanjing 210003,China
关键词:
零样本学习; 嵌入模型; 属性学习; 深度神经网络; 公共空间; 词向量
Keywords:
zero-shot learning; embedding model; attribute learning; deep neural network; co-space; word2vec
分类号:
TP181
DOI:
10.3969/ j. issn.1673-629X.2018.11.010
文献标志码:
A
摘要:
零样本学习(ZSL)是一种特殊的机器学习问题,目的是在测试阶段识别出训练集中未曾出现的类别样本并进行分类。目前主流技术手段有两种:一种是基于属性学习,一种是基于词嵌入模型。 两种方法各有优缺点。属性学习可以看作人工特征标记,分类效果往往取决于人工设定的属性好坏。而利用文本特征(词向量)的词嵌入模型无需人为参与也可得到不错的特征以替代属性。 文中提出了一种基于公共空间嵌入的深度零样本学习方法。通过图像和文本分别建立深度神经网络并连接两个网络,在顶层学习一个联合嵌入的公共空间。 基于深度学习端到端模型的设计可以同时利用属性特征和文本特征实现图像的零样本学习。 实验结果表明,该方法达到了较好的识别效果。
Abstract:
Zero-shot learning (ZSL) is a special machine learning problem and aims to classify the samples without appearing in the training set at test stage. At present,there are two main technical methods:one is based on attribute learning and the other is based word embedding. They have their own advantages and disadvantages. Attribute learning can be regarded as artificial feature and its classification effect often depends on the quality of artificial mark. Word embedding model using text feature (word2vector) can get great feature to replace attribute without human participation. In this paper,we propose a deep zero-shot learning based on co-space embedding. The deep neural networks are established respectively by image and text and connected to learn a jointly embedding co-space at the top. The design of the end-to-end model based on deep learning can use both attribute and text feature in image ZSL. Experiment shows that this method achieves better recognition results.

相似文献/References:

[1]刘 帅,黄 刚,戴晓峰,等.一种融合生成对抗网络的零样本图像分类方法[J].计算机技术与发展,2022,32(07):87.[doi:10. 3969 / j. issn. 1673-629X. 2022. 07. 015]
　LIU Shuai,HUANG Gang,DAI Xiao-feng,et al.A Zero-shot Classification Based on Generative Adversarial Network[J].,2022,32(11):87.[doi:10. 3969 / j. issn. 1673-629X. 2022. 07. 015]
[2]宁园园,张素兰,陈 飞.基于双注意力机制的零样本建筑图像分类方法[J].计算机技术与发展,2023,33(10):35.[doi:10. 3969 / j. issn. 1673-629X. 2023. 10. 006]
　NING Yuan-yuan,ZHANG Su-lan,CHEN Fei.Zero-shot Architectural Image Classification Method Based on Dual Attention Mechanism[J].,2023,33(11):35.[doi:10. 3969 / j. issn. 1673-629X. 2023. 10. 006]

更新日期/Last Update: 2018-11-10