Non-negative Matrix Factorization
in [Mathematica]
|
From: 孙挺 on 30 Nov 2009 06:15 Hi community, I'm dealing with a problem which needs NMF(Non-negative Matrix Factorization). I've looked up the document along with Mathematica, but nothing found. And I've tried to implement one function, however, the efficiency is too poor in comparison with built-in SingularValueDecompostion[]. Is there any implement of NMF? Thank you all! sunt
From: Daniel Lichtblau on 1 Dec 2009 04:13 Sunt wrote: > Hi community, > I'm dealing with a problem which needs NMF(Non-negative Matrix > Factorization). > I've looked up the document along with Mathematica, but nothing found. > And I've tried to implement one function, however, the efficiency is too > poor in comparison with built-in SingularValueDecompostion[]. > Is there any implement of NMF? > > Thank you all! > > sunt If dimensions are modest, could try method posted at URL below. http://forums.wolfram.com/mathgroup/archive/2009/Mar/msg00739.html One advantage to that method is you might be able to get reasonable speed if you work with low rank approximations so there are not too many variables. Daniel Lichtblau Wolfram Research
From: dh on 1 Dec 2009 04:20 孙挺 wrote: > Hi community, > I'm dealing with a problem which needs NMF(Non-negative Matrix > Factorization). > I've looked up the document along with Mathematica, but nothing found. > And I've tried to implement one function, however, the efficiency is too > poor in comparison with built-in SingularValueDecompostion[]. > Is there any implement of NMF? > > Thank you all! > > sunt > > Hi, your problem is ill defined. One may factor a non negative matrix in infinite many ways. It all depends on further qualities of the matrix and the aim one want to achieve. Is there a reason why is "Eigensystem" or "SingularValueDecompostion", although they do not care about non-negativity, does not suit you? Daniel
|
Pages: 1 Prev: RootReduce ver 7 Next: Simple Spreadsheet within Mathematica? |