Doing NMF in R with missing values (NA) in the matrix

Question

I am looking for a package/if-possible-relatively-off-the-shelf solution for NMF (non-negative matrix factorization) in R, that can handle missing values (NA) and not consider them as 0.

The goal is in fact to estimate these missing values through the product of the factorization, for a simple recommender system.

The NMF CRAN-package is great, but doesn't seems able to do that (neither can its much more recent continuation off-CRAN), and I couldn't find suitable alternative packages...

Answer 1

NLLM (https://cran.r-project.org/web/packages/NNLM/index.html) does the missing value imputation via non-negative matrix factorization in R.

Set the missing values of the matrix as NA. Please see the vignette for exact details.