--

Hi Eric, after looking to your github repo I have some thoughts:

- How do you inject the user and item features to the model training, is it somehow hardcoded? I see some private methods taking use of the in the RankFM class, but not an explicit way for so.

- What if we don't have *features* for all the users and items implied in the interactions?

--

--

Augusto de Nevrezé
Augusto de Nevrezé

Written by Augusto de Nevrezé

I write about Data Science, AI, ML & DL. I’m electronics engineer. My motto: “Per Aspera Ad Astra”. Follow me in twtr @augusto_dn