This is our implementation of the paper:
Xin Zhou, Aixin Sun, Yong Liu, Jie Zhang, Chunyan Miao. SelfCF: A Simple Framework for Self-supervised Collaborative Filtering http://arxiv.org/abs/2107.03019.
We present a simple framework that specialized for collaborative filtering problem. The framework enables learning of latent representations of users and items without Negative Samples.
We augment the output embeddings generated from backbone networks instead of the input user-item ids. For output embedding augmentation, we propose three techniques:
- Historical embedding.
- Embedding dropout.
- Edge pruning.
For details, please refer to the paper.
- Standard data preprocessing.
- Unified data splitting with global-timeline.
- Standard evaluation protocols.
- Posterior recommendation results.
Download from Google Drive: MovieLens/MOOC/Amazon-Vedio-Game
python 3.6
pytorch 1.8
PyYAML 0.1.7
pandas 0.24
numpy 1.19
We would like to give thanks to the following repos:
RecBole
BUIR