This is the PyTorch implementation by @HelloElwin for MAERec proposed in the paper Graph Masked Autoencoder for Sequential Recommendation published in SIGIR'23 by Yaowen Ye, Lianghao Xia, and Chao Huang.
MAERec is a simple yet effective graph masked autoencoder that adaptively and dynamically distills global item transitional information for self-supervised augmentation through a novel adaptive transition path masking strategy. It naturally addresses the data scarcity and noise perturbation problems in sequential recommendation scenarios and avoids issues in most contrastive learning-based methods.
We suggest the following environment for running MAERec:
python==3.8.13
pytorch==1.12.1
numpy==1.18.1
Please first unzip the desired dataset in the dataset folder, and then run
- Amazon Books:
python main.py --data books
- Amazon Toys:
python main.py --data toys
- Retailrocket:
python main.py --data retailrocket
More explanation of model hyper-parameters can be found here.
If you find this work helpful to your research, please kindly consider citing our paper:
@inproceedings{ye2023graph,
title={Graph Masked Autoencoder for Sequential Recommendation},
author={Ye, Yaowen and Xia, Lianghao and Huang, Chao},
booktitle={Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR'23), July 23-27, 2023, Taipei, Taiwa},
year={2023}
}