Institutional Repository of Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
Deep unsupervised learning with consistent inference of latent representations | |
Chang, Jianlong1,2; Wang, Lingfeng1; Meng, Gaofeng1; Xiang, Shiming1,2; Pan, Chunhong1 | |
发表期刊 | PATTERN RECOGNITION |
2018-05-01 | |
卷号 | 77期号:5页码:438-453 |
文章类型 | Article |
摘要 | Utilizing unlabeled data to train deep neural networks (DNNs) is a crucial but challenging task. In this paper, we propose an end-to-end approach to tackle this problem with consistent inference of latent representations. Specifically, each unlabeled data point is considered as a seed to generate a set of latent labeled data points by adding various random disturbances or transformations. Under the expectation maximization framework, DNNs can be trained in an unsupervised way by minimizing the distances between the data points with the same latent representations. Furthermore, several variants of our approach can be derived by applying regularized and sparse constraints during optimization. Theoretically, the convergence of the proposed method and its variants are fully analyzed. Experimental results show that the proposed approach can significantly improve the performance on various tasks, including image classification and clustering. Such results also indicate that our method can guide DNNs to learn more invariant feature representations in comparison with traditional unsupervised methods. (C) 2017 Elsevier Ltd. All rights reserved. |
关键词 | Deep Unsupervised Learning Consistent Inference Of Latent Representations |
WOS标题词 | Science & Technology ; Technology |
DOI | 10.1016/j.patcog.2017.10.022 |
关键词[WOS] | NEURAL-NETWORKS ; AUTO-ENCODERS ; RECOGNITION ; CLASSIFICATION ; ALGORITHM |
收录类别 | SCI |
语种 | 英语 |
项目资助者 | National Natural Science Foundation of China (NSFC)(91646207 ; Beijing Nature Science Foundation(4162064) ; Youth Innovation Promotion Association CAS ; 61403376 ; 61370039 ; 91338202) |
WOS研究方向 | Computer Science ; Engineering |
WOS类目 | Computer Science, Artificial Intelligence ; Engineering, Electrical & Electronic |
WOS记录号 | WOS:000426222800033 |
引用统计 | |
文献类型 | 期刊论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/20364 |
专题 | 模式识别国家重点实验室_先进时空数据分析与学习 |
作者单位 | 1.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100049, Peoples R China |
第一作者单位 | 模式识别国家重点实验室 |
推荐引用方式 GB/T 7714 | Chang, Jianlong,Wang, Lingfeng,Meng, Gaofeng,et al. Deep unsupervised learning with consistent inference of latent representations[J]. PATTERN RECOGNITION,2018,77(5):438-453. |
APA | Chang, Jianlong,Wang, Lingfeng,Meng, Gaofeng,Xiang, Shiming,&Pan, Chunhong.(2018).Deep unsupervised learning with consistent inference of latent representations.PATTERN RECOGNITION,77(5),438-453. |
MLA | Chang, Jianlong,et al."Deep unsupervised learning with consistent inference of latent representations".PATTERN RECOGNITION 77.5(2018):438-453. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
chang-pr-2018.pdf(2654KB) | 期刊论文 | 作者接受稿 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论