As someone who is very used to thinking in terms of matrices and who just went through the adjustment of translating matlab-like code to numpy, I found the current matrix module to be confusing.  It's poor integration with the rest of numpy/scipy (in particular, scipy.optimize.fmin_cg) made it more difficult to use than it was worth.  I'd rather have "matrix" and/or "matrix multiplication" sections of the documentation explain how to do typical, basic matrix operations with nparray, dot, T, and arr[None,:].  I think a matrix class would still be worthwhile for findability, but it should simply serve as documentation for how to do matrix stuff with nparray.

Cheers,

Jason

--
Jason Rennie
Research Scientist, ITA Software
617-714-2645
http://www.itasoftware.com/