search

spotify-research / cosernn

Code for the paper "Contextual and Sequential User Embeddings for Large-Scale Music Recommendation".
Apache License 2.0
33 stars 8 forks source link

how the feature "time elapsed since previous session" and "number of tracks in previous session" used? #6

Open animesh-wynk opened 3 months ago

animesh-wynk commented 3 months ago

hi!

for simplicity, lets call the feature "time elapsed since previous session" f1 and "number of tracks in previous session" f2

doubts related to f1:

  1. is f1 in hours or minutes or is it in some other unit?
  2. how are you normalising this feature?

doubts related to f12:

  1. it is mentioned in the paper that in your dataset, avg number of songs in user sessions is 10, and that you have only considered the first 10 songs in a session to compute the session embedding. then how is the f2 calculated? is it the actual number of songs in the previous song?
  2. how is this feature normalised?

thanks a lot in advance : )

animesh-wynk commented 3 months ago

hi @lucasmaystre ! can you please help me out here? : )

lucasmaystre commented 3 months ago

Hi @animesh-wynk, unfortunately I don't remember these implementation details. The code is the authoritative resource for these questions, hopefully you should be able to find the answers to there in the code.