PreFact: Knowledge Propagation Regulating Network Toward Preferred Facts for Knowledge-aware Recommendation

This is our implementation for the paper: PreFact: Knowledge Propagation Regulating Network Toward Preferred Facts for Knowledge-aware Recommendation

Environment Settings

• torch==2.7.1
• python==3.13.4
• pykeen==1.11.1
• networkx==3.3
• pandas==2.3.0
• prettytable==3.16.0
• setproctitle==1.3.6
• torch-scatter==2.1.2

Example to run the codes.

1. Config dataset and model_name in KGE.py, then run it to get pretrained KGE embeddings. This is done by python package pykeen.

2. Run the following command to train PreFact on Last-FM dateset with GPU 0:

   python main.py --dataset last-f --gpu_id 0
   

3. After training, you can check the log files in ./logs

Parameter Tuning

All the parameters are in ./utils/parser.py