Readme

This repo contains Python codes for Disentangled Representation Learning for Recommendation. IEEE Transactions on Pattern Analysis and Machine Intelligence 2022.

The paper can be downloaded from the official website or in the docs directory.

Files

• main.py: main code to train, test and visualize
• model.py: MultiDAE, MultiVAE, DisenVAE(MacridVAE), DisenEVAE(SEM-MacridVAE)
• data.py: load dataset and split into train, val_tr, val_te, test_tr, test_te
• RecomData
  • ml-latest-small: sample dataset
    • prep.py: pre-process data
    • ratings.txt: rating actions file
    • embed.npy: image features file
  • prep.py: further process data
• run: save log and model

Run

### part 1
./DecomData/<dataset>

### part 2
./DecomData
python prep.py <dataset>

### part 3
./
python main.py --data <dataset> --model <model> --device <cuda>


### Take ml-latest-small dataset as an example: 
./DecomData/ml-latest-small:  python prep.py
./DecomData:                  python prep.py ml-latest-small
# In this repository, the above two steps have been completed, and the results have been saved in the prep directory.
# It is worth noting that we keep users who have at least five rating actions, instead of fifteen in our paper.
# If you would like to keep users who have at least fifteen rating actions, modify the code in line 52 in ./RecomData/prep.py.
./:                           python main.py --data ml-latest-small --model DisenEVAE --device cuda:0

Cite

@article{wang2022disentangled,
  title={Disentangled representation learning for recommendation},
  author={Wang, Xin and Chen, Hong and Zhou, Yuwei and Ma, Jianxin and Zhu, Wenwu},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  volume={45},
  number={1},
  pages={408--424},
  year={2022},
  publisher={IEEE}
}