A Unified Convergence Analysis of the Multiplicative Update Algorithm for Regularized Nonnegative Matrix Factorization

被引:22
作者
Zhao, Renbo [1 ,2 ,3 ]
Tan, Vincent Y. F. [1 ,3 ]
机构
[1] NUS, Dept Elect & Comp Engn, Singapore 117583, Singapore
[2] NUS, Dept Ind Syst Engn & Management, Singapore 117576, Singapore
[3] NUS, Dept Math, Singapore 119076, Singapore
关键词
Nonnegative matrix factorization; multiplicative update algorithm; convergence analysis; nonconvex optimization; stationary points; ALPHA-BETA; OPTIMIZATION; DIVERGENCES;
D O I
10.1109/TSP.2017.2757914
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
The multiplicative update (MU) algorithm has been extensively used to estimate the basis and coefficient matrices in nonnegative matrix factorization (NMF) problems under a wide range of divergences and regularizers. However, theoretical convergence guarantees have only been derived for a few special divergences without regularization. In this work, we provide a conceptually simple, self-contained, and unified proof for the convergence of the MU algorithm applied on NMF with a wide range of divergences and regularizers. Our main result shows the sequence of iterates (i.e., pairs of basis and coefficient matrices) produced by the MU algorithm converges to the set of stationary points of the nonconvex NMF optimization problem. Our proof strategy has the potential to open up new avenues for analyzing similar problems in machine learning and signal processing.
引用
收藏
页码:129 / 138
页数:10
相关论文
共 56 条
  • [1] [Anonymous], ARXIV08104225
  • [2] [Anonymous], 1999, Athena scientific Belmont
  • [3] [Anonymous], 2010, P 2010 SIAM INT C DA
  • [4] Arora S, 2012, STOC'12: PROCEEDINGS OF THE 2012 ACM SYMPOSIUM ON THEORY OF COMPUTING, P145
  • [5] Stability Analysis of Multiplicative Update Algorithms and Application to Nonnegative Matrix Factorization
    Badeau, Roland
    Bertin, Nancy
    Vincent, Emmanuel
    [J]. IEEE TRANSACTIONS ON NEURAL NETWORKS, 2010, 21 (12): : 1869 - 1881
  • [6] Burden R., 2016, NUMERICAL ANAL
  • [7] Graph Regularized Nonnegative Matrix Factorization for Data Representation
    Cai, Deng
    He, Xiaofei
    Han, Jiawei
    Huang, Thomas S.
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2011, 33 (08) : 1548 - 1560
  • [8] ON TENSORS, SPARSITY, AND NONNEGATIVE FACTORIZATIONS
    Chi, Eric C.
    Kolda, Tamara G.
    [J]. SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 2012, 33 (04) : 1272 - 1299
  • [9] Non-negative matrix factorization with α-divergence
    Cichocki, Andrzej
    Lee, Hyekyoung
    Kim, Yong-Deok
    Choi, Seungjin
    [J]. PATTERN RECOGNITION LETTERS, 2008, 29 (09) : 1433 - 1440
  • [10] Generalized Alpha-Beta Divergences and Their Application to Robust Nonnegative Matrix Factorization
    Cichocki, Andrzej
    Cruces, Sergio
    Amari, Shun-ichi
    [J]. ENTROPY, 2011, 13 (01) : 134 - 170