An implementation of the Singular Value Decomposition (SVD) of a product of matrices (implicitly, without forming the product). I've written it in Matlab before, and now ported it to Julia.

With some testing and utility and examples. Possibly useful for data mining (the first Netflix price was won using the SVD).

• The original paper by Golub, Solna, van Dooren
• My report on the original implementation in Matlab