I'm unsure if I should post this on Stack overflow or Mathematics but I thought this question fits into algorithm studies so here I am.
I've written an algorithm to calculate the pseudoinverse of a given matrix. I've done this using a naïve way using the definition A+ = (A*A)-1A *.
I've read more into this and found that a common way to calculate it more stable and efficiëntly is calculating the SVD of A first. (I do this using the Jama package)
However. The naïve method proves to be faster (in time) on small matrices.
A quick question. I believe the SVD method has a better complexity than the naïve way. But can you give me an idea on how big the matrix has to be so the SVD method is actually faster in terms of time?