Those that work in .NET languages (e.g. C#, VB, F#) may want to have a look at the Meta.Numerics library. Information is available here (http://www.meta-numerics.net/) and source code and bits are available here (http://metanumerics.codeplex.com/).

Meta.Numerics is not a simple port of NR to C#. It many ways it is much less than NR: it does not (yet) do integration, interpolation, or Fourier analysis. But in some ways it is more: it's functionality is packaged in a coherent object model consistent with .NET best practices. If your primary interests are in

matrix operations (including eigenvalues and eigenvectors of arbitrary square matrices)
functions of mathematical physics (including complex arguments of the Gamma and error functions)
statistical analysis (including contingency analysis, fitting model parameters using experiemntal data, and statistical tests)

then Meta.Numerics will interest you.

Meta.Numerics is released under a permissive license (MS-PL) and is maintained by a Microsoft employee who works with the .NET team.