Is SVD included in MathNet.Numerics x86?

Question

I need to calculate the Singular Value Decomposition of a Dense matrix but it doesn't seem to be included in the package I'm using: MathNet.Numerics x86 v2.4.0.26 downloaded from Nuget package manager.

I am referencing this question Svd recomposition..

The syntax included in the answer linked is:

 var m = DenseMatrix.OfArray(new double[,] {
   { 3, 0, 0, 0, 0 },
   { 0, 2, 4, 0, 0 },
   { 0, 4, 5, -4, 5 },
   { 0, 0, -4, -8, 12},
   { 0, 0, 5, 12, -5 }});


  var svd = m.Svd(true);  //The method Svd() doesn't seem to be available in v2.4 

  svd.U() * svd.W() * svd.VT()

Nor is svd() listed in the documentation.

I'm looking for a simple example of generating a SVD of a DenseMatrix using the MathNet.Numerics x86 library.

Unfortunately the method inverse() is just returning NaN so I'm hoping to approximate the inverse using singular value decomposition.

Answer 1

From the documentation I found classes for singular value decomposition

MathNet.Numerics.LinearAlgebra. Double/Single/Generic .Factorization.Svd are abstract classes.

MathNet.Numerics.LinearAlgebra. Double/Single/Generic .Factorization.DenseSvd are implementations. Pass your matrix to the constructor, the results are available via members.

Answer 2

Svd() used to be an extension method in v2, which unfortunately is only available if you include the right namespace. In your case, adding the following line should do the trick:

using MathNet.Numerics.LinearAlgebra.Double;

This is one of the areas that have been greatly simplified in the upcoming v3 release, where this is a proper method. I suggest you have a look at one of the recent v3 packages (as of writing this e.g. the v3.0.0-alpha7).