This repository contains TensorFlow codes and datasets for the paper:

Lianghao Xia, Chao Huang, Chuxu Zhang (2022). Self-Supervised Hypergraph Transformer for Recommender Systems, Paper in arXiv, Paper in ACM. In KDD'22, Washington DC, USA, August 14-18, 2022.

This paper proposed a hypergraph transformer model named SHT to tackle the data sparsity and noise problem in collaborative filtering. It combines user/item id embedding with the topology-aware embedding given by GCNs as the local view, and learns the latent global relations in a transformer-like hypergraph encoder. The local embeddings are then denoised by the global view using an augmented solidity differentiation task. Comparing to our previous work HCCF which is on hypergraph contrastive learning for CF, this paper proposes the novel hypergraph transformer architecture with the self-augmented solidity differentiation task, and highlights the strength of generative self-supervised learning.

If you want to use our codes and datasets in your research, please cite:

@inproceedings{sht2022,
  author    = {Xia, Lianghao and
               Huang, Chao and
	       Zhang, Chuxu},
  title     = {Self-Supervised Hypergraph Transformer for Recommender Systems},
  booktitle = {Proceedings of the 28th {ACM} {SIGKDD} Conference on
               Knowledge Discovery and Data Mining, {KDD} 2022, 
	       Washington DC, USA, August 14-18, 2022.},
  year      = {2022},
}

The codes of SHT are implemented and tested under the following development environment:

• python=3.7.0
• tensorflow=1.15.0
• numpy=1.21.6
• scipy=1.7.3

Three datasets are adopted to evaluate SHT: Yelp, Gowalla, and Tmall. The user-item pair $(u_i, v_j)$ in the adjacent matrix is set as 1, if user $u_i$ has rated item $v_j$ in Yelp, or if user $u_i$ has check in venue $v_j$ in Gowalla, or if user $u_i$ has purchased item $v_j$ in Tmall. We filtered out users and items with too few interactions.

Please unzip the datasets first. Also you need to create the Models/ directory. The following command lines start training and testing on the three datasets, respectively, which also specify the hyperparameter settings for the reported results in the paper. Training and testing logs for trained models are contained in the History/ directory.

• Yelp

python .\labcode_hop.py --data yelp --reg 1e-2 --ssl_reg 1e-5 --mult 1e2 --edgeSampRate 0.1 --epoch 1

• Gowalla

python .\labcode_hop.py --data gowalla --reg 1e-2 --ssl_reg 1e-5 --mult 1e1 --epoch 150 --edgeSampRate 0.1 --save_path gowalla_repeat

• Tmall

python .\labcode_hop.py --data tmall --reg 1e-2 --ssl_reg 1e-5 --mult 1e1 --edgeSampRate 0.1 --epoch 150 --save_path tmall_repeat

Important arguments:

• reg: This is the weight for weight-decay regularization. Empirically recommended tuning range is {1e-2, 1e-3, 1e-4, 1e-5}.
• ssl_reg: This is the weight for the solidity prediction loss of self-supervised learning task. The value is tuned from {1e-3, 1e-4, 1e-5, 1e-6, 1e-7}.
• mult: This hyperparameter is to emplify the ssl loss for better performance, which is tuned from {16, 64, 1e1, 1e2, 1e3}.
• edgeSampRate: This parameter determines the ratio of edges to conduct the solidity differentiation task on. It should be balanced to consider both model performance and training efficiency.

This research work is supported by the research grants from the Department of Computer Science & Musketeers Foundation Institute of Data Science at the University of Hong Kong.