OnlineNMF.jl Documentation

Overview

OnlineNMF.jl binarizes CSV file, summarizes the information of data matrix and, performs some online-NMF functions for extreamly large scale matrix in an out-of-core manner without loading whole data on memory space.

Online NMF methods are performed as following two steps.

• Step.1 Binarization : We asssume that the data is matrix filled in integar count value and saved as comma-separated CSV, Matrix Market (MM), or Binary COO (BinCOO) file. Using OnlinePCA package, These files are converted to Julia binary file by csv2bin, mm2bin or bincoo2bin, respectively. This step extremely accelerates I/O speed.
• Step.2 Online NMF : nmf and dnmf can be perfomed against the binary file generated by csv2bin. sparse_nmf and sparse_dnmf can be perfomed against the binary file generated by mm2bin. bincoo_nmf and bincoo_dnmf can be perfomed against the binary file generated by bincoo2bin.

All programs are available as Julia API (OnlineNMF.jl (Julia API)) and command line tool (OnlineNMF.jl (Command line tool)).

Reference

• Multiplicative Update (MU)
  • Alpha-divergence: Cichocki, A. et al., 2008
    • Alpha=2 : Pearson divergence-based NMF
    • Alpha=0 or 1 : Kullback–Leibler (KL) divergence-based NMF
    • Alpha=0.5 : Hellinger divergence-based NMF
  • Beta-divergence: Févotte, C. et al., 2011, Nakano, M. et al., 2010
    • Beta=2 : Euclidean distance-based NMF with Gaussian distribution
    • Beta=1 : Kullback–Leibler divergence-based NMF with Poisson distribution
    • Beta=0 : Itakura-Saito divergence-based NMF with Gamma distribution
• Discretized Non-negative Matrix Factorization (DNMF): Koki Tsuyuzaki, 2023