Sketching for Simultaneously Sparse and Low-Rank Covariance Matrices

被引:0
|
作者
Bahmani, Sohail [1 ]
Romberg, Justin [1 ]
机构
[1] Georgia Inst Technol, Sch Elect & Comp Engn, Atlanta, GA 30332 USA
基金
美国国家科学基金会;
关键词
RECOVERY;
D O I
暂无
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
We introduce a technique for estimating a structured covariance matrix from observations of a random vector which have been sketched. Each observed random vector x(t) is reduced to a single number by taking its inner product against one of a number of pre-selected vector a(l). These observations are used to form estimates of linear observations of the covariance matrix Sigma, which is assumed to be simultaneously sparse and low-rank. We show that if the sketching vectors a(l) have a special structure, then we can use straightforward two-stage algorithm that exploits this structure. We show that the estimate is accurate when the number of sketches is proportional to the maximum of the rank times the number of significant rows/columns of Sigma. Moreover, our algorithm takes direct advantage of the low-rank structure of Sigma by only manipulating matrices that are far smaller than the original covariance matrix.
引用
收藏
页码:357 / 360
页数:4
相关论文
共 50 条
  • [11] Sparse Representation of a Polytope and Recovery of Sparse Signals and Low-Rank Matrices
    Cai, T. Tony
    Zhang, Anru
    IEEE TRANSACTIONS ON INFORMATION THEORY, 2014, 60 (01) : 122 - 132
  • [12] Rate Optimal Denoising of Simultaneously Sparse and Low Rank Matrices
    Yang, Dan
    Ma, Zongming
    Buja, Andreas
    JOURNAL OF MACHINE LEARNING RESEARCH, 2016, 17
  • [13] Recovering low-rank and sparse components of matrices for object detection
    Zhang, Hanling
    Liu, Liangliang
    ELECTRONICS LETTERS, 2013, 49 (02) : 109 - 110
  • [14] EXACT RECOVERY OF LOW-RANK PLUS COMPRESSED SPARSE MATRICES
    Mardani, Morteza
    Mateos, Gonzalo
    Giannakis, Georgios B.
    2012 IEEE STATISTICAL SIGNAL PROCESSING WORKSHOP (SSP), 2012, : 49 - 52
  • [15] A subspace clustering algorithm based on simultaneously sparse and low-rank representation
    Liu, Xiaolan
    Yi, Miao
    Han, Le
    Deng, Xue
    JOURNAL OF INTELLIGENT & FUZZY SYSTEMS, 2017, 33 (01) : 621 - 633
  • [16] Sparse and Low-Rank Decomposition of Covariance Matrix for Efficient DOA Estimation
    Chen, Yong
    Wang, Fang
    Wan, Jianwei
    Xu, Ke
    2017 IEEE 9TH INTERNATIONAL CONFERENCE ON COMMUNICATION SOFTWARE AND NETWORKS (ICCSN), 2017, : 957 - 961
  • [17] Optimal Denoising of Simultaneously Sparse and Low Rank Matrices in High Dimensions
    Buja, Andreas
    Ma, Zongming
    Yang, Dan
    2013 51ST ANNUAL ALLERTON CONFERENCE ON COMMUNICATION, CONTROL, AND COMPUTING (ALLERTON), 2013, : 445 - 447
  • [18] Sketching Sparse Low-Rank Matrices With Near-Optimal Sample- and Time-Complexity Using Message Passing
    Liu, Xiaoqi
    Venkataramanan, Ramji
    IEEE TRANSACTIONS ON INFORMATION THEORY, 2023, 69 (09) : 6071 - 6097
  • [19] Recovering Low-Rank and Sparse Matrices via Robust Bilateral Factorization
    Shang, Fanhua
    Liu, Yuanyuan
    Cheng, James
    Cheng, Hong
    2014 IEEE INTERNATIONAL CONFERENCE ON DATA MINING (ICDM), 2014, : 965 - 970
  • [20] AN ALGEBRAIC MULTILEVEL PRECONDITIONER WITH LOW-RANK CORRECTIONS FOR SPARSE SYMMETRIC MATRICES
    Xi, Yuanzhe
    Li, Ruipeng
    Saad, Yousef
    SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2016, 37 (01) : 235 - 259