search

quora / qmf

A fast and scalable C++ library for implicit-feedback matrix factorization models
Apache License 2.0
463 stars 97 forks source link

MovieLens performance #19

Open bkj opened 6 years ago

bkj commented 6 years ago

Hi All --

Has anyone ever benchmarks this package on the MovieLens datasets? I'm looking at this for the first time and trying to figure out what hyperparamers need to be tuned to get good performance.

Thanks! ~ Ben

EDIT: Related -- if anyone has a pointer to a properly parameterized example of using this library, that'd also be super helpful.

albietz commented 6 years ago

Hi,

I'm not sure anyone has tried QMF on MovieLens, but iirc MovieLens consists of data with explicit feedback (i.e. ratings, while QMF deals with implicit data, like "clicks")? Either way, if you do try it out, let us know how it does!

Alberto

bkj commented 6 years ago

I think it's fine to binarize MovieLens and treat it as implicit feedback -- the number of movies reviewed per user is pretty small. Then you try to optimize eg. the p@k over some held out data split.

~ Ben

albietz commented 6 years ago

Fair -- feel free to post your findings here if you try that out.

Alberto