Non-convex low-rank representation combined with rank-one matrix sum for subspace clustering

被引:0
|
作者
Liu, Xiaofang [2 ]
Wang, Jun [1 ]
Cheng, Dansong [1 ]
Shi, Daming [3 ]
Zhang, Yongqiang [1 ]
机构
[1] Harbin Inst Technol, Sch Comp Sci & Technol, 92 West Dazhi St, Harbin, Peoples R China
[2] Harbin Inst Technol, Sch Elect Engn & Automat, 92 West Dazhi St, Harbin, Peoples R China
[3] Shenzhen Univ, Coll Comp & Software, 3688 Nanhai Ave, Shenzhen, Peoples R China
基金
中国国家自然科学基金;
关键词
Subspace clustering; Non-convex low-rank representation; Block coordinate descent; Rank-one matrix; DECISION-MAKING; ALGORITHM;
D O I
10.1007/s00500-020-04865-0
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Exploring the multiple subspace structures of data such as low-rank representation is effective in subspace clustering. Non-convex low-rank representation (NLRR) via matrix factorization is one of the state-of-the-art techniques for subspace clustering. However, NLRR cannot scale to problems with large n (number of samples) as it requires either the inversion of an nxn linear system. To address this issue, we propose a novel approach, NLRR++, which reformulates NLRR as a sum of rank-one components, and apply a column-wise block coordinate descent to update each component iteratively. NLRR++ reduces the time complexity per iteration from O(n3) and the memory complexity from O(n2), where m is the dimensionality and d is the target rank (usually dMUCH LESS-THANmMUCH LESS-THANn). Our experimental results on simulations and real datasets have shown the efficiency and effectiveness of NLRR++. We demonstrate that NLRR++ is not only much faster than NLRR, but also scalable to large datasets such as the ImageNet dataset with 120K samples.
引用
收藏
页码:15317 / 15326
页数:10
相关论文
共 50 条
  • [21] A Non-convex One-Pass Framework for Generalized Factorization Machine and Rank-One Matrix Sensing
    Lin, Ming
    Ye, Jieping
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 29 (NIPS 2016), 2016, 29
  • [22] Rank Consistency Induced Multiview Subspace Clustering via Low-Rank Matrix Factorization
    Guo, Jipeng
    Sun, Yanfeng
    Gao, Junbin
    Hu, Yongli
    Yin, Baocai
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2022, 33 (07) : 3157 - 3170
  • [23] Adaptive iterative hard thresholding for low-rank matrix recovery and rank-one measurements
    Xia, Yu
    Zhou, Likai
    JOURNAL OF COMPLEXITY, 2023, 76
  • [24] One-step Low-Rank Representation for Clustering
    Fu, Zhiqiang
    Zhao, Yao
    Chang, Dongxia
    Wang, Yiming
    Wen, Jie
    Zhang, Xingxing
    Guo, Guodong
    PROCEEDINGS OF THE 30TH ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA, MM 2022, 2022, : 2220 - 2228
  • [25] Two Rank Approximations for Low-Rank Based Subspace Clustering
    Xu, Fei
    Peng, Chong
    Hu, Yunhong
    He, Guoping
    2017 10TH INTERNATIONAL CONGRESS ON IMAGE AND SIGNAL PROCESSING, BIOMEDICAL ENGINEERING AND INFORMATICS (CISP-BMEI), 2017,
  • [26] Low-Rank Positive Semidefinite Matrix Recovery From Corrupted Rank-One Measurements
    Li, Yuanxin
    Sun, Yue
    Chi, Yuejie
    IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2017, 65 (02) : 397 - 408
  • [27] Non-Convex Low-Rank Approximation for Image Denoising and Deblurring
    Lei, Yang
    Song, Zhanjie
    Song, Qiwei
    IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS, 2016, E99D (05): : 1364 - 1374
  • [28] Non-Convex Low-Rank Matrix Recovery from Corrupted Random Linear Measurements
    Li, Yuanxin
    Chi, Yuejie
    Zhang, Huishuai
    Liang, Yingbin
    2017 INTERNATIONAL CONFERENCE ON SAMPLING THEORY AND APPLICATIONS (SAMPTA), 2017, : 134 - 137
  • [29] A non-convex optimization framework for large-scale low-rank matrix factorization
    Hafshejani, Sajad Fathi
    Vahidian, Saeed
    Moaberfard, Zahra
    Lin, Bill
    MACHINE LEARNING WITH APPLICATIONS, 2022, 10
  • [30] Adaptive low-rank kernel block diagonal representation subspace clustering
    Liu, Maoshan
    Wang, Yan
    Sun, Jun
    Ji, Zhicheng
    APPLIED INTELLIGENCE, 2022, 52 (02) : 2301 - 2316