PyTorch implementations of some Recommendation System papers
- scipy==1.1.0
- networkx==2.1
- pandas==0.23.4
- numpy==1.15.1
- joblib==1.0.1
- torch==1.8.0
cd runs
python run_bprmf.pyYelp: https://www.kaggle.com/yelp-dataset/yelp-dataset
We followed the Leave-One-Out evaluation strategy in SASRec. Specifically, for each user, we randomly sample 100 negative items and rank these items with the ground-truth item. HR and NDCG are estimated based on the ranking results.
| Test Result | Recall@10 | NDCG@10 |
|---|---|---|
| BPRMF | 76.51±0.26 | 55.77±0.16 |
| NeuMF | 79.35±0.12 | 59.06±0.24 |
| GRec | 81.55±0.17 | 55.74±0.18 |
| DGRec | 86.57±0.18 | 63.55±0.26 |
| SocialMF | 76.27±0.28 | 53.42±0.21 |
| SoRec | 81.45±0.04 | 58.15±0.07 |
| LightGCN | 84.39±0.07 | 60.80±0.19 |
| SASRec | 81.66±0.08 | 57.21±0.37 |
| ASASRec | 84.53±0.04 | 60.53±0.09 |
| TransRec | 80.19±0.20 | 64.00±0.15 |
This repo contains the following models:
- BPRMF: https://arxiv.org/abs/1205.2618
- NeuMF: https://arxiv.org/abs/1708.05031
- GRec: https://dl.acm.org/doi/abs/10.1145/3308558.3313488
- DGRec: https://dl.acm.org/doi/abs/10.1145/3289600.3290989
- SocialMF: https://dl.acm.org/doi/abs/10.1145/1864708.1864736
- SoRec: https://dl.acm.org/doi/abs/10.1145/1458082.1458205
- LightGCN: https://dl.acm.org/doi/abs/10.1145/3397271.3401063
- SASRec: https://arxiv.org/abs/1808.09781
- ASASREec: https://dl.acm.org/doi/10.1145/3397271.3401264
- TransRec: https://dl.acm.org/doi/abs/10.1145/3109859.3109882