Skip to content

Latest commit

 

History

History
57 lines (45 loc) · 2.01 KB

README.md

File metadata and controls

57 lines (45 loc) · 2.01 KB
Raw

REST

The PyTorch implementation of paper REST: Debiased Social Recommendation via Reconstructing Exposure Strategies.

Usage

Download dataset

Download datasets: Ciao, Epinions, Yelp

Then unzip them into the directory datasets and run preprocess_yelp.py.

└── datasets
    ├── Ciao
    │   ├── rating_with_timestamp.mat
    │   ├── trust.mat
    ├── Epinions
    │   ├── rating_with_timestamp.mat
    │   ├── trust.mat
    ├── Yelp
    │   ├── yelp_academic_dataset_review.json
    │   ├── yelp_academic_dataset_user.json
    │   ├── noiso_reid_u2uir.npz
    │   ├── ...

Run

    python ./run_rate/run_rest_rate_ciao.py

During the training, we can obtain some logs and model-checkpoints in the directory logs and saved_models,

Results

Model Ciao RMSE Epinions RMSE Yelp RMSE
PMF 1.1936±0.0019 1.2755±0.0022 1.2454±0.0011
NeuMF 0.9828±0.0022 1.0838±0.0015 1.1958±0.0005
MultiVAE 1.1908±0.0014 1.2104±0.0039 1.2944±0.0020
RecVAE 1.1787±0.0022 1.1946±0.0038 1.2385±0.0014
CausE 1.0003±0.0013 1.0705±0.0013 1.2039±0.0015
CVIB-MF 1.2001±0.0011 1.2477±0.0003 1.3189±0.0024
CVIB-NCF 1.0462±0.0013 1.2477±0.0003 1.3613±0.0043
MACR-MF 1.1859±0.0030 1.2364±0.0031 1.2344±0.0004
DecRS 0.9875±0.0033 1.0617±0.0033 -
GraphRec 0.9743±0.0021 1.0567±0.0019 1.1968±0.0017
NGCF 1.0135±0.0010 1.1286±0.0017 1.2231±0.0017
LightGCN 1.1919±0.0014 1.2025±0.0005 1.2444±0.0019
REST 0.9635±0.0009 1.0413±0.0007 1.1733±0.0006

Detailed results can be found in the paper.