MGCN: Multi-View Graph Convolutional Network for Multimedia Recommendation

Introduction

This is the Pytorch implementation for our MM 2023 paper:

MM 2023. Penghang Yu, Zhiyi Tan, Guanming Lu, Bing-Kun Bao(2023). Multi-View Graph Convolutional Network for Multimedia Recommendation

News! Our Latest work

[AAAI 2025 Oral]

We propose a Principal Graph Learning (PGL) method for multimedia recommendation, achieving state-of-the-art performance.

Paper Link Code Link

Compared to MGCN, PGL delivers over 7% performance improvement.

We appreciate your interest and welcome feedback on our latest work!

Enviroment Requirement

• python 3.8
• Pytorch 1.12

Dataset

We provide three processed datasets: Baby, Sports and Clothing.

Download from Google Drive: Baby/Sports/Clothing

Training

cd ./src
python main.py

Performance Comparison

Citing MGCN

If you find MGCN useful in your research, please consider citing our paper.

@article{yu2023multi,
  title={Multi-View Graph Convolutional Network for Multimedia Recommendation},
  author={Yu, Penghang and Tan, Zhiyi and Lu, Guanming and Bao, Bing-Kun},
  booktitle={Proceedings of the 31st ACM International Conference on Multimedia},
  pages = {6576–6585},
  year={2023}
}

The code is released for academic research use only. For commercial use, please contact Penghang Yu.

Acknowledgement

The structure of this code is based on MMRec. Thank for their work.