search

maciejkula / spotlight

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

Does it make sense to build a sequential model? #153

Open sak360 opened 5 years ago

sak360 commented 5 years ago

Just stumbled upon this project and it look great!

I am trying to conceptually understand what type of recommender to build. The sequence in my data set cannot be assumed (think of it as more like a bucket of items) and only contains implicit data....does it still make sense to try building sequential models or should I just stick with an implicit MF approach?

maciejkula commented 5 years ago

One advantage of sequential models is that they can immediately ingest new interactions: you don't need to retrain the model to update your user representations with new actions. If the order of interactions isn't important to you, try the pooling representation.