Is there good library to do nonnegative matrix factorization (NMF) fast?

Question

I have a sparse matrix whose shape is 570000*3000. I tried nima to do NMF (using the default nmf method, and set max_iter to 65). However, I found nimfa very slow. Have anyone used a faster library to do NMF?

Answer 1

I have used libNMF before. It's written in C and is very fast. There is a paper documenting the algorithm and code.

The paper also lists several alternative packages for NMF (in bunch of different languages (which I have copied here for future reference).

• The Mathworks [3, 33]
  • Matlab
  • http://www.mathworks.com/access/helpdesk/help/toolbox/stats/nnmf.
• Cemgil [5]
  • Matlab
  • http://www.cmpe.boun.edu.tr/~cemgil/bnmf
• Cichocki et al. [6]
  • Matlab
  • http://www.bsp.brain.riken.jp/ICALAB/nmflab.
• Cichocki et al. [7]
  • Matlab
  • http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0470746661.html
• Hansen et al. [14]
  • Matlab
  • http://isp.imm.dtu.dk/toolbox/nmf/index.html
• Hoyer [16]
  • Matlab
  • http://www.cs.helsinki.fi/u/phoyer/software.html
• Kim et al. [19]
  • Matlab
  • http://userweb.cs.utexas.edu/users/dmkim/Source/software/nnma/index.html
• Lin [25]
  • Matlab/Python
  • http://www.csie.ntu.edu.tw/~cjlin/nmf/index.html
• Schmidt et al. [30]
  • Matlab
  • http://mikkelschmidt.dk/index.php?id=2
• Gaujoux [10]
  • R
  • http://cran.r-project.org/web/packages/NMF/index.html
• Liu [26]
  • R
  • http://cran.r-project.org/web/packages/NMFN/index.html
• Battenberg et al. 2
  • Python
  • http://www.eecs.berkeley.edu/~ericb
• Schmitt et al. [31]
  • Python
  • http://www.procoders.net/?p=409
• Dhillon et al. [8]
  • C++
  • http://www.kyb.mpg.de/bs/people/suvrit/work/progs/nnma.html
• Greene et al. [13]
  • C++
  • http://mlg.ucd.ie/nmf
• Pathak et al. [28]
  • C++
  • http://www.insight-journal.org/browse/publication/152
• Wang et al. [34]
  • C++
  • http://www.biomedcentral.com/1471-2105/7/175

Disclaimer: I have not tried any of these other packages (aside from MATLAB's).