Documentation: Difference between this and "regular" SVD

[BradKML]

As Wikipedia had suggested, Funk et. al. is not using the regular form of SVD with three variable, but rather a modified form of NMF with two variables but faster convergence and speed.

With reference to the Funk:

• https://www.wikiwand.com/en/Matrix_factorization_(recommender_systems)#/Funk_MF

With reference to SVD:

• https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.TruncatedSVD.html
• https://numpy.org/doc/stable/reference/generated/numpy.linalg.svd.html

Would it be good to explain how the optimization differs from NMF?

[gbolmier]

Hi @BrandonKMLee, Thank you for taking the time to drop a comment 😃

So far the README.md is the only documentation of the project. I would prefer to keep it short, concise, and not go too far into the theoretical details. The important part of the optimization process is already covered and a software documentation isn't supposed to explain much about the mathematical part. Nevertheless, I think it makes sense to add references for people interested in more background; I'll clean up the README.md and add some in the next days 👍

[BradKML]

I think I might have asked the same question in an SVD repo that is slightly more relevant. ZWMiller/svdRec#1

[gbolmier]

Added a References section in README.md (cf. corresponding commit)