search

maciejkula / spotlight

Deep recommender models using PyTorch.
MIT License
2.97k stars 421 forks source link

Visual Bayesian Personalized Ranking #158

Open abkds opened 5 years ago

abkds commented 5 years ago

I am proposing a new feature. We can extend the BPR module according to this paper, this includes features about the items itself. It is generic in the sense that the developer, along with the interactions, also gives the system some sort of embeddings about the items and the system then tries to learn the latent factors.

aldrinc commented 5 years ago

@abkds Take a look at https://github.com/maciejkula/spotlight/issues/144 where Maciej has a basic implementation of item & user features in FM. It's not 100% plug and play but part of the fun is figuring how everything works. 👍

abkds commented 5 years ago

Cool that helps, actually I kind of implemented vBPR for one of my own projects, so I was thinking of bringing it here, but if it's already here, I will check it out. Thanks 💯