Skip to content

Rule-Guided Graph Neural Networks for Recommender Systems, ISWC 2020

License

Notifications You must be signed in to change notification settings

nju-websoft/RGRec

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Rule-Guided Graph Neural Networks for Recommender Systems

Contributions Welcome License language-python3 made-with-Pytorch

To alleviate the cold start problem caused by collaborative filtering in recommender systems, knowledge graphs (KGs) are increasingly employed by many methods as auxiliary resources. However, existing work incorporated with KGs cannot capture the explicit long-range semantics between users and items meanwhile consider various connectivity between items. In this paper, we propose RGRec, which combines rule learning and graph neural networks (GNNs) for recommendation. RGRec first maps items to corresponding entities in KGs and adds users as new entities. Then, it automatically learns rules to model the explicit long-range semantics, and captures the connectivity between entities by aggregation to better encode various information. We show the effectiveness of RGRec on three real-world datasets. Particularly, the combination of rule learning and GNNs achieves substantial improvement compared to methods only using either of them.

Getting Started

Package Description

src/
├── aggregators.py: a deprecated module whose methods have been moved into RGRec_model.py
├── ALogger.py: log management module
├── Args.py: global argument module
├── auc_f1_main.py: train module, and a part of test module which tests auc and f1
├── Graph.py: data structure module 
├── hit_ndcg_main.py: test module which involves testing hits@k and ndcg@k
├── pra_model.py: rule weights pre-training module
├── RGRec_model.py: the core module to implement RGRec
├── Sundries.py: provide a simple function

Dependencies

  • Python 3.x (tested on Python 3.6)
  • Pytorch 1.x (tested on Pytorch 1.3)
  • Numpy
  • Scipy
  • Sklearn

Usage

  1. Set your arguments in Args.py.
  2. Run pra_model.py to get the pre-trained weights of rules.
  3. Run auc_f1_main.py to train the model and obtain the test results about auc and f1.
  4. Run hit_ndcg_main.py to obtain the test results about hits@k and ndcg@k.

Experiment

We conduct experiments on three real-world datasets: Last.FM, MovieLens-1M and Dianping-Food. Currently, our project only provides MovieLens-1M, other datasets can be downloaded here.

The details of experiments can be found in the corresponding section of the paper.

License

This project is licensed under the GPL License - see the LICENSE file for details.

Miscellaneous

During the construction of RGRec, we obtain much inspiration from KGCN, and we feel a great honor to make an innovation in the foundation of this work.

Citation

If you use this work or code, please kindly cite the following paper:

@inproceedings{RGRec,
  author    = {Xinze Lyu and Guangyao Li and Jiacheng Huang and Wei Hu},
  title     = {Rule-Guided Graph Neural Networks for Recommender Systems},
  booktitle = {ISWC},
  year      = {2020},
}

Contacts

If you have any questions, please feel free to contact Guangyao Li, we will reply it as soon as possible.

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •  

Languages