AI helps you reading Science

AI generates interpretation videos

AI extracts and analyses the key points of the paper to generate videos automatically


pub
Go Generating

AI Traceability

AI parses the academic lineage of this thesis


Master Reading Tree
Generate MRT

AI Insight

AI extracts a summary of this paper


Weibo:
We study one challenging issue in multi-view data clustering

Partially View-aligned Clustering

NIPS 2020, (2020)

Cited by: 0|Views83
EI
Full Text
Bibtex
Weibo

Abstract

In this paper, we study one challenging issue in multi-view data clustering. To be specific, for two data matrices X(1) and X(2) corresponding to two views, we do not assume that X(1) and X(2) are fully aligned in row-wise. Instead, we assume that only a small portion of the matrices has established the correspondence in advance. Such a p...More

Code:

Data:

0
Introduction
  • As one of the most important unsupervised technologies, data clustering has attracted much attention in recent years [22, 16, 6, 34].
  • Most existing multi-view clustering (MVC) approaches jointly learn a common representation to bridge the gap among different views and achieve clustering using the common representation.
  • The success of such a learning paradigm highly relies on a “well-established” dataset which has to satisfy two assumptions: 1) completeness of data: It requires that all examples appear in all views.
  • With the above two assumptions, the correlation and correspondence of multi-view data are available, making learning the common representation and clustering possible
Highlights
  • As one of the most important unsupervised technologies, data clustering has attracted much attention in recent years [22, 16, 6, 34]
  • To tackle the partially view-aligned problem in MVC, we propose a novel neural network which simultaneously aligns a given partially aligned dataset in a latent space and learns the common representation across different views
  • Based on the above discussions, we aim to develop a neural network that could establish the correspondence of a given dataset using a differentiable surrogate of the Hungarian algorithm and simultaneously learn the common representation for different views by implicitly using the alignment information
  • A challenging problem in multi-view clustering, namely the partially view-aligned problem, is studied for the first time. The solution to this problem could alleviate intensive labor for fully-aligned data collection. To solve this challenging problem, we propose a novel multi-view clustering method termed as partially view-aligned clustering consists of a differentiable alignment module and a representation learning module
  • The alignment module is a differentiable surrogate of the non-differentiable Hungarian algorithm, which could establish the correspondence of two views
  • While there will be important impacts resulting from the use of partially view-aligned clustering (PVC) in general, here we focus on the impact of using our method to address the partially view-aligned problem which is widely faced by the real-world applications
Methods
  • The authors compare the PVC with nine multi-view clustering approaches including: canonically correlated analysis (CCA)[29], kernel canonically correlated analysis (KCCA)[2], deep canonically correlated analysis (DCCA) [1], deep canonically correlated autoencoders (DCCAE) [30], Multi-View Clustering via Deep Matrix Factorization(MvC-DMF) [40], latent multi-view subspace clustering (LMSC) [36], self-weighted multi-view clustering (SwMC) [23], binary multi-view clustering (BMVC) [39], and Autoencoder in Autoencoder Networks (AE2-Nets) [37].
  • For MvC-DMF, the authors seek the optimal β and γ from (0.1, 1, 10, 100) as suggested
Results
  • PVC achieves the best result when 80% rather than 100% of data are with ground-truth correspondence.
Conclusion
  • A challenging problem in multi-view clustering, namely the partially view-aligned problem, is studied for the first time
  • The solution to this problem could alleviate intensive labor for fully-aligned data collection.
  • To solve this challenging problem, the authors propose a novel multi-view clustering method termed as partially view-aligned clustering consists of a differentiable alignment module and a representation learning module.
  • It is still unknown how to handle fully unaligned data and the data which simultaneously encounters the missing views and unaligned data problems
Tables
  • Table1: Clustering performance comparison on four challenging datasets
Download tables as Excel
Related work
  • Multi-view clustering methods aim to exploit the diverse and complementary information contained in different views [32], which could be roughly classified into three categories based on different formulations of view-specific similarity and cross-view consistency. Namely, multi-view canonical correlation clustering which utilizes the correlation among different views [29, 1, 30]; multi-view matrix decomposition clustering which exploits the mutual information based on the matrix factorization technology and then perform the clustering on the learned matrix with low-rank constraint [40]; and multi-view subspace clustering which jointly perform the subspace learning with view-specific similarity and learn the common space with cross-modal consistency[23, 36, 15]. Although the above approaches have achieved promising results in multi-view clustering, they highly rely on the aforementioned two assumptions, i.e., “completeness of data” and “correspondence of views”. Thus they are limited to handling the partially data-missing problem (PDP) and partially view-aligned problem (PVP) well.

    Recently, there are some deep methods have been proposed to solve PDP [13, 28, 4, 9, 18, 17]. The basic idea of these works is to utilize the remained information in the available views to predict the t = 8s t = 11s
Funding
  • This work was supported in part by NFSC under Grant U19A2081, 61625204, and 61836006; in part by the Fundamental Research Funds for the Central Universities under Grant YJ201949; in part by the Fund of Sichuan University Tomorrow Advancing Life; and in part by A*STAR AME Programmatic under Grant A18A1b0045
Study subjects and analysis
widely-used multi-view datasets with the comparisons of nine multi-view clustering approaches: 4
After that, the view-specific representations are simply concatenated as the common representation which is further used for clustering by k-means like the traditional fashion [30, 25, 37]. In this section, we evaluate the proposed PVC method on four widely-used multi-view datasets with the comparisons of nine multi-view clustering approaches. We implement PVC in PyTorch and carry all evaluations on a standard Ubuntu-18.04 OS with an NVIDIA 2080Ti GPU

popular multi-view datasets: 4
4.1 Experiment Setting. We carry our experiments on four popular multi-view datasets including: Caltech101-20 [15, 25] which consists of 2,386 images of 20 subjects with two handcrafted features as two views. Reuters [8] which is a subset of the Reuters database

samples: 3000
Reuters [8] which is a subset of the Reuters database. It consists of 3,000 samples from 6 classes, using German and Spanish as two views. Scene-15[5] which consists of 4,485 images distributed over 15 scene categories with two views

datasets: 4
4.2 Comparison with State of The Arts. To evaluate the effectiveness of PVC on the partially view-aligned data, we first construct the partially view-aligned data from the above four datasets. For Caltech101-20, Reuters and Scene-15, we randomly split them into two partitions ({A(v), U(v)}m v=1) with the equal size

Reference
  • Galen Andrew, Raman Arora, Jeff Bilmes, and Karen Livescu. Deep canonical correlation analysis. In Proc. Int. Conf. Mach. Learn., pages 1247–1255, 2013.
    Google ScholarLocate open access versionFindings
  • Francis R Bach and Michael I Jordan. Kernel independent component analysis. J. Mach. Learn. Res., 3(Jul):1–48, 2002.
    Google ScholarLocate open access versionFindings
  • James P Boyle and Richard L Dykstra. A method for finding projections onto the intersection of convex sets in hilbert spaces. In Advances in order restricted statistical inference, pages 28–47.
    Google ScholarLocate open access versionFindings
  • Lei Cai, Zhengyang Wang, Hongyang Gao, Dinggang Shen, and Shuiwang Ji. Deep adversarial learning for multi-modality missing data completion. In Proc. ACM SIGKDD Int. Conf. Knowl. Discov. & Data Min., pages 1158–1166, 2018.
    Google ScholarLocate open access versionFindings
  • Li Fei-Fei and Pietro Perona. A bayesian hierarchical model for learning natural scene categories. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., volume 2, pages 524–531. IEEE, 2005.
    Google ScholarLocate open access versionFindings
  • Chenping Hou, Feiping Nie, Dongyun Yi, and Dacheng Tao. Discriminative embedded clustering: A framework for grouping high-dimensional data. IEEE Trans. Neural Netw. Learn. Syst., 26(6):1287–1299, 2014.
    Google ScholarLocate open access versionFindings
  • Xin Huang, Yuxin Peng, and Mingkuan Yuan. Mhtn: Modal-adversarial hybrid transfer network for cross-modal retrieval. IEEE Trans. Cybern., 2018.
    Google ScholarLocate open access versionFindings
  • Zhenyu Huang, Joey Tianyi Zhou, Xi Peng, Changqing Zhang, Hongyuan Zhu, and Jiancheng Lv. Multi-view spectral clustering network. In Proc. Int. Joint Conf. Artifi. Intell., pages 2563–2569, 10–16 Aug 2019.
    Google ScholarLocate open access versionFindings
  • Yangbangyan Jiang, Qianqian Xu, Zhiyong Yang, Xiaochun Cao, and Qingming Huang. Dm2c: Deep mixed-modal clustering. In Proc. Adv. Neural Inf. Process. Syst., pages 5880–5890, 2019.
    Google ScholarLocate open access versionFindings
  • Harold W Kuhn. The hungarian method for the assignment problem. Naval research logistics quarterly, 2(1-2):83–97, 1955.
    Google ScholarLocate open access versionFindings
  • Christoph H Lampert and Oliver Krömer. Weakly-paired maximum covariance analysis for multimodal dimensionality reduction and transfer learning. In Proc. Eur. Conf. Comput. Vis., pages 566–579.
    Google ScholarLocate open access versionFindings
  • Hendrik W Lenstra Jr. Integer programming with a fixed number of variables. Mathematics of operations research, 8(4):538–548, 1983.
    Google ScholarLocate open access versionFindings
  • Shao-Yuan Li, Yuan Jiang, and Zhi-Hua Zhou. Partial multi-view clustering. In Proc. AAAI Conf. Artif. Intell., 2014.
    Google ScholarLocate open access versionFindings
  • Sheng Li, Ming Shao, and Yun Fu. Cross-view projective dictionary learning for person re-identification. In Proc. Int. Joint Conf. Artifi. Intell., 2015.
    Google ScholarLocate open access versionFindings
  • Yeqing Li, Feiping Nie, Heng Huang, and Junzhou Huang. Large-scale multi-view spectral clustering via bipartite graph. In Proc. Conf. AAAI Artif. Intell., pages 2750–2756, 2015.
    Google ScholarLocate open access versionFindings
  • Guangcan Liu, Zhouchen Lin, Shuicheng Yan, Ju Sun, Yong Yu, and Yi Ma. Robust recovery of subspace structures by low-rank representation. IEEE Trans. Pattern Anal. Mach. Intell., 35(1):171–184, 2012.
    Google ScholarLocate open access versionFindings
  • Xinwang Liu, Lei Wang, Xinzhong Zhu, Miaomiao Li, En Zhu, Tongliang Liu, Li Liu, Yong Dou, and Jianping Yin. Absent multiple kernel learning algorithms. IEEE Trans. Pattern Anal. Mach. Intell., 42(6):1303–1316, 2019.
    Google ScholarLocate open access versionFindings
  • Xinwang Liu, Xinzhong Zhu, Miaomiao Li, Lei Wang, Chang Tang, Jianping Yin, Dinggang Shen, Huaimin Wang, and Wen Gao. Late fusion incomplete multi-view clustering. IEEE Trans. Pattern Anal. Mach. Intell., 41(10):2410–2423, 2018.
    Google ScholarLocate open access versionFindings
  • Canyi Lu, Shuicheng Yan, and Zhouchen Lin. Convex sparse spectral clustering: Single-view to multi-view. IEEE Trans. Image Process., 25(6):2833–2843, 2016.
    Google ScholarLocate open access versionFindings
  • Sobhan Miryoosefi, Kianté Brantley, Hal Daume III, Miro Dudik, and Robert E Schapire. Reinforcement learning with convex constraints. In Adv. Neural Inf. Process. Syst., pages 14093–14102, 2019.
    Google ScholarLocate open access versionFindings
  • Kathleen L Mosier, Linda J Skitka, Susan Heers, and Mark Burdick. Automation bias: Decision making and performance in high-tech cockpits. The International journal of aviation psychology, 8(1):47–63, 1998.
    Google ScholarLocate open access versionFindings
  • Andrew Y Ng, Michael I Jordan, and Yair Weiss. On spectral clustering: Analysis and an algorithm. In Proc. Adv. Neural Inf. Process. Syst., pages 849–856, 2002.
    Google ScholarLocate open access versionFindings
  • Feiping Nie, Jing Li, and Xuelong Li. Self-weighted multiview clustering with multiple graphs. In Proc. Int. Joint Conf. Artifi. Intell., pages 2564–2570, 2017.
    Google ScholarLocate open access versionFindings
  • François-Pierre Paty and Marco Cuturi. Subspace robust wasserstein distances. arXiv preprint arXiv:1901.08949, 2019.
    Findings
  • Xi Peng, Zhenyu Huang, Jiancheng Lv, Hongyuan Zhu, and Joey Tianyi Zhou. Comic: Multiview clustering without parameter selection. In Proc. Int. Conf. Mach. Learn., pages 5092–5101, 2019.
    Google ScholarLocate open access versionFindings
  • Richard Sinkhorn. A relationship between arbitrary positive matrices and doubly stochastic matrices. The annals of mathematical statistics, 35(2):876–879, 1964.
    Google ScholarLocate open access versionFindings
  • Zhiqiang Tao, Hongfu Liu, Sheng Li, Zhengming Ding, and Yun Fu. From ensemble clustering to multi-view clustering. In Proc. Int. Joint Conf. Artifi. Intell., 2017.
    Google ScholarLocate open access versionFindings
  • Luan Tran, Xiaoming Liu, Jiayu Zhou, and Rong Jin. Missing modalities imputation via cascaded residual autoencoder. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 1405–1414, 2017.
    Google ScholarLocate open access versionFindings
  • Alexei Vinokourov, Nello Cristianini, and John Shawe-Taylor. Inferring a semantic representation of text via cross-language correlation analysis. In Proc. Adv. Neural Inf. Process. Syst., pages 1497–1504, 2003.
    Google ScholarLocate open access versionFindings
  • Weiran Wang, Raman Arora, Karen Livescu, and Jeff Bilmes. On deep multi-view representation learning. In Proc. Int. Conf. Mach. Learn., pages 1083–1092, 2015.
    Google ScholarLocate open access versionFindings
  • Cai Xu, Ziyu Guan, Wei Zhao, Hongchang Wu, Yunfei Niu, and Beilei Ling. Adversarial incomplete multi-view clustering. In Proc. Int. Joint Conf. Artifi. Intell., pages 3933–3939, 2019.
    Google ScholarLocate open access versionFindings
  • Chang Xu, Dacheng Tao, and Chao Xu. A survey on multi-view learning. arXiv preprint arXiv:1304.5634, 2013.
    Findings
  • Chang Xu, Dacheng Tao, and Chao Xu. Multi-view intact space learning. IEEE Trans. Pattern Anal. Mach. Intell., 37(12):2531–2544, 2015.
    Google ScholarLocate open access versionFindings
  • Yingzhen Yang, Jiashi Feng, Nebojsa Jojic, Jianchao Yang, and Thomas S Huang. Subspace learning by 0-induced sparsity. Int. J. Comput. Vis., 126(10):1138–1156, 2018.
    Google ScholarLocate open access versionFindings
  • Xiaohui Zeng, Renjie Liao, Li Gu, Yuwen Xiong, Sanja Fidler, and Raquel Urtasun. Dmm-net: Differentiable mask-matching network for video object segmentation. In Proc. IEEE Int. Conf. Comput. Vis., October 2019.
    Google ScholarLocate open access versionFindings
  • Changqing Zhang, Qinghua Hu, Huazhu Fu, Pengfei Zhu, and Xiaochun Cao. Latent multi-view subspace clustering. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 4279–4287, 2017.
    Google ScholarLocate open access versionFindings
  • Changqing Zhang, Yeqing Liu, and Huazhu Fu. Ae2-nets: Autoencoder in autoencoder networks. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., pages 2577–2585, 2019.
    Google ScholarLocate open access versionFindings
  • Xianchao Zhang, Linlin Zong, Xinyue Liu, and Hong Yu. Constrained nmf-based multi-view clustering on unmapped data. In Proc. AAAI Conf. Artif. Intell., 2015.
    Google ScholarLocate open access versionFindings
  • Zheng Zhang, Li Liu, Fumin Shen, Heng Tao Shen, and Ling Shao. Binary multi-view clustering. IEEE Trans. Pattern Anal. Mach. Intell., 2018.
    Google ScholarLocate open access versionFindings
  • Handong Zhao, Zhengming Ding, and Yun Fu. Multi-view clustering via deep matrix factorization. In Proc. AAAI Conf. Artif. Intell., pages 2921–2927, 2017.
    Google ScholarLocate open access versionFindings
  • Wei Zhao, Cai Xu, Ziyu Guan, and Ying Liu. Multiview concept learning via deep matrix factorization. IEEE Trans. Neural Netw. Learn. Syst., 2020.
    Google ScholarLocate open access versionFindings
Your rating :
0

 

Tags
Comments
数据免责声明
页面数据均来自互联网公开来源、合作出版商和通过AI技术自动分析结果,我们不对页面数据的有效性、准确性、正确性、可靠性、完整性和及时性做出任何承诺和保证。若有疑问,可以通过电子邮件方式联系我们:report@aminer.cn
小科