How to include additional information in movie lens collaborative filtering example?

(WG) #1

Reading through the documentation, I noticed that the MovieLens datasets include much more than just userIds, movieIds, ratings and titles.

My question is: How/Where would I incorporate this additional information in building either a NN or using the dot product+bias method presented in the lesson 4 notebook?

For example,

  1. Lets say I wanted to code and include the genres identified for each movie, where each movieId is associated to a vector where each element represents a genre (1 if movie part of that genre, 0 otherwise). Where would this be plugged into the NN?

  2. Or lets say I wanted to include a bias term into the NN (like to dot product+bias example) … where would this get plugged in?

(Cristian) #2

Hey @wgpubs, that would be really interesting to me as well.
Did you discover something useful regarding this?


(WG) #3


Lesson 7 covers how to do this. In particular, pay attention to the “Multi-input” section in the lesson 7 notebook.