Awesome
<img src="https://i.ibb.co/54vTYzk/ssl-logo.png" alt="ssl-logo" border="0"> <p float="left"><img src="https://img.shields.io/badge/python-v3.7+-red"> <img src="https://img.shields.io/badge/pytorch-v1.7+-blue"> <img src="https://img.shields.io/badge/tensorflow-v1.14+-green"> <br>SELFRec is a Python framework for self-supervised recommendation (SSR) which integrates commonly used datasets and metrics, and implements many state-of-the-art SSR models. SELFRec has a lightweight architecture and provides user-friendly interfaces. It can facilitate model implementation and evaluation. <br> Founder and principal contributor: @Coder-Yu @xiaxin1998 <br>
This repo is released with our survey paper on self-supervised learning for recommender systems. We organized a tutorial on self-supervised recommendation at WWW'22. Visit the tutorial page for more information.
Supported by:<br> Prof. Hongzhi Yin, The University of Queensland, Australia, h.yin1@uq.edu.au <br> Prof. Shazia Sadiq, ARC Training Centre for Information Resilience (CIRES), University of Queensland, Australia
<h2>Architecture<h2> <img src="https://raw.githubusercontent.com/Coder-Yu/SELFRec/main/selfrec.jpg" alt="ssl-logo" border="0" style="width:600px"> <h2>Features</h2> <ul> <li><b>Fast execution</b>: SELFRec is compatible with Python 3.9+, Tensorflow 1.14+ (optional), and PyTorch 1.8+ and powered by GPUs. We also optimize the time-consuming item ranking procedure, drastically reducing ranking time to seconds. </li> <li><b>Easy configuration</b>: SELFRec provides simple and high-level interfaces, making it easy to add new SSR models in a plug-and-play fashion.</li> <li><b>Highly Modularized</b>: SELFRec is divided into multiple discrete and independent modules. This design decouples model design from other procedures, allowing users to focus on the logic of their method and streamlining development.</li> <li><b>SSR-Specific</b>: SELFRec is designed specifically for SSR. It provides specific modules and interfaces for rapid development of data augmentation and self-supervised tasks.</li> </ul> <h2>How to Use</h2> <ol> <li>Execute <b>pip install -r requirements.txt</b> under the SELFRec directory</li> <li>Configure the xx.yaml file in ./conf . (xx is the name of the model you want to run)</li> <li>Run main.py and choose the model you want to run.</li> </ol> <h2>Implemented Models</h2> <table class="table table-hover table-bordered"> <tr> <th>Model</th> <th>Paper</th> <th>Type</th> <th>Code</th> </tr> <tr> <td scope="row">SASRec</td> <td>Kang et al. <a href="https://cseweb.ucsd.edu/~jmcauley/pdfs/icdm18.pdf" target="_blank">Self-Attentive Sequential Recommendation</a>, ICDM'18. </td> <td>Sequential</d> <td>PyTorch</d> </tr> <tr> <td scope="row">CL4SRec</td> <td>Xie et al. <a href="https://arxiv.org/abs/2010.14395" target="_blank">Contrastive Learning for Sequential Recommendation</a>, ICDE'22. </td> <td>Sequential</d> <td>PyTorch</d> </tr> <tr> <td scope="row">BERT4Rec</td> <td>Sun et al. <a href="https://dl.acm.org/doi/pdf/10.1145/3357384.3357895" target="_blank">BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformer</a>, CIKM'19. </td> <td>Sequential</d> <td>PyTorch</d> </tr> <table> <table class="table table-hover table-bordered"> <tr> <th>Model</th> <th>Paper</th> <th>Type</th> <th>Code</th> </tr> <tr> <td scope="row">XSimGCL</td> <td>Yu et al. <a href="https://arxiv.org/abs/2209.02544" target="_blank">XSimGCL: Towards Extremely Simple Graph Contrastive Learning for Recommendation</a>, TKDE'23. </td> <td>Graph + CL</d> <td>PyTorch</d> </tr> <tr> <td scope="row">SimGCL</td> <td>Yu et al. <a href="https://arxiv.org/abs/2112.08679" target="_blank">Are Graph Augmentations Necessary? Simple Graph Contrastive Learning for Recommendation</a>, SIGIR'22. </td> <td>Graph + CL</d> <td>PyTorch</d> </tr> <tr> <td scope="row">DirectAU</td> <td>Wang et al. <a href="https://arxiv.org/abs/2206.12811" target="_blank">Towards Representation Alignment and Uniformity in Collaborative Filtering</a>, KDD'22. </td> <td>Graph</d> <td>PyTorch</d> </tr> <tr> <td scope="row">NCL</td> <td>Lin et al. <a href="https://arxiv.org/abs/2202.06200" target="_blank">Improving Graph Collaborative Filtering with Neighborhood-enriched Contrastive Learning</a>, WWW'22. </td> <td>Graph + CL</d> <td>PyTorch</d> </tr> <tr> <td scope="row">MixGCF</td> <td>Huang et al. <a href="https://keg.cs.tsinghua.edu.cn/jietang/publications/KDD21-Huang-et-al-MixGCF.pdf" target="_blank">MixGCF: An Improved Training Method for Graph Neural Network-based Recommender Systems</a>, KDD'21. </td> <td>Graph + DA</d> <td>PyTorch</d> </tr> <tr> <td scope="row">MHCN</td> <td>Yu et al. <a href="https://dl.acm.org/doi/abs/10.1145/3442381.3449844" target="_blank">Self-Supervised Multi-Channel Hypergraph Convolutional Network for Social Recommendation</a>, WWW'21. </td> <td>Graph + CL</d> <td>TensorFlow</d> </tr> <tr> <td scope="row">SGL</td> <td>Wu et al. <a href="https://dl.acm.org/doi/10.1145/3404835.3462862" target="_blank">Self-supervised Graph Learning for Recommendation</a>, SIGIR'21. </td> <td>Graph + CL</d> <td>TensorFlow & Torch</d> </tr> <tr> <td scope="row">SEPT</td> <td>Yu et al. <a href="https://arxiv.org/abs/2106.03569" target="_blank">Socially-Aware Self-supervised Tri-Training for Recommendation</a>, KDD'21. </td> <td>Graph + CL</d> <td>TensorFlow</d> </tr> <tr> <td scope="row">BUIR</td> <td>Lee et al. <a href="https://arxiv.org/abs/2105.06323" target="_blank">Bootstrapping User and Item Representations for One-Class Collaborative Filtering</a>, SIGIR'21. </td> <td>Graph + DA</d> <td>PyTorch</d> </tr> <tr> <td scope="row">SSL4Rec</td> <td>Yao et al. <a href="https://dl.acm.org/doi/abs/10.1145/3459637.3481952" target="_blank">Self-supervised Learning for Large-scale Item Recommendations</a>, CIKM'21. </td> <td>Graph + CL</d> <td>PyTorch</d> </tr> <tr> <td scope="row">SelfCF</td> <td>Zhou et al. <a href="https://arxiv.org/abs/2107.03019" target="_blank">SelfCF: A Simple Framework for Self-supervised Collaborative Filtering</a>, arXiv'21. </td> <td>Graph + DA</d> <td>PyTorch</d> </tr> <tr> <td scope="row">LightGCN</td> <td>He et al. <a href="https://dl.acm.org/doi/10.1145/3397271.3401063" target="_blank">LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation</a>, SIGIR'20. </td> <td>Graph</d> <td>PyTorch</d> </tr> <tr> <td scope="row">MF</td> <td>Yehuda et al. <a href="https://ieeexplore.ieee.org/abstract/document/5197422" target="_blank">Matrix Factorization Techniques for Recommender Systems</a>, IEEE Computer'09. </td> <td>Graph</d> <td>PyTorch</d> </tr> </table> * CL is short for contrastive learning (including data augmentation); DA is short for data augmentation only <h2>Leaderboard</h2> The results are obtained on the dataset of <b>Yelp2018</b>. We performed grid search for the best hyperparameters. <br> General hyperparameter settings are: batch_size: 2048, emb_size: 64, learning rate: 0.001, L2 reg: 0.0001. <br><br>| Model | Recall@20 | NDCG@20 | Hyperparameter settings |
|---|---|---|---|
| MF | 0.0543 | 0.0445 | |
| LightGCN | 0.0639 | 0.0525 | layer=3 |
| NCL | 0.0670 | 0.0562 | layer=3, ssl_reg=1e-6, proto_reg=1e-7, tau=0.05, hyper_layers=1, alpha=1.5, num_clusters=2000 |
| SGL | 0.0675 | 0.0555 | λ=0.1, ρ=0.1, tau=0.2 layer=3 |
| MixGCF | 0.0691 | 0.0577 | layer=3, n_nes=64, layer=3 |
| DirectAU | 0.0695 | 0.0583 | 𝛾=2, layer=3 |
| SimGCL | 0.0721 | 0.0601 | λ=0.5, eps=0.1, tau=0.2, layer=3 |
| XSimGCL | 0.0723 | 0.0604 | λ=0.2, eps=0.2, l∗=1 tau=0.15 layer=3 |
- Create a .conf file for your model in the directory named conf.
- Make your model inherit the proper base class.
- Reimplement the following functions.
- build(), train(), save(), predict()
- Register your model in main.py.
@article{yu2023self,
title={Self-supervised learning for recommender systems: A survey},
author={Yu, Junliang and Yin, Hongzhi and Xia, Xin and Chen, Tong and Li, Jundong and Huang, Zi},
journal={IEEE Transactions on Knowledge and Data Engineering},
year={2023},
publisher={IEEE}
}