search

lyst / lightfm

A Python implementation of LightFM, a hybrid recommendation algorithm.
Apache License 2.0
4.66k stars 679 forks source link

Implementing Incremental Training and Feedback loop using Lightfm #705

Open abadugu13 opened 5 months ago

abadugu13 commented 5 months ago

I have successfully trained a lightfm model based on user features, item features and interactions. Now I want to incrementally train the model to cover two aspects:

  1. Addition of new items/users and their interactions
  2. Update model based on user feedback(preferably both positive and negative) of the provided recommendations

For the first case, I can see the fit_partial function in the Dataset and Model which seem to be the way, but I'm not sure. For the second case, I couldn't find anything in the documentation, so I am not sure how to proceed. Any directions would be appreciated