Skip to content

Latest commit

 

History

History
23 lines (16 loc) · 1.03 KB

readme.md

File metadata and controls

23 lines (16 loc) · 1.03 KB
Raw

LightGCN on Pytorch

This is a implementation of LightGCN (Paper in arXiv) neural net from SIGIR 2020

Supported datasets:

Use prepare_<dataset_name>_dataset.py for download and splitting by time

Supported models:

  • iALS is matrix factorization model from implicit open-source library
  • TopNModel recommends top items from all user feedback
  • TopNPersonalized recommends top items from unique user feedback
  • TopNNearestModel recommends nearest by last user location items (domain-specific for geo features)
  • LightGCN
  • Catboost fitting with LogLoss/YetiRank and ranking candidates

Training:

Main script is train.py which trains model from MODEL setting in config.yaml file

Also there is fit_catboost.py script which trains catboost ranking model