AdvDrop

Overview

Code of "General Debiasing for Graph-based Collaborative Filtering via Adversarial Graph Dropout"

Run the Code

• We provide implementation for various baselines presented in the paper.

• To run the code, first install Cython via pip and run the following command to install tools used in evaluation:

python setup.py build_ext --inplace

AdvDrop Training

Coat:

python run_AdvDrop.py --modeltype AdvDrop --dataset Coat.new --n_layers 2 --neg_sample 1 --saveID yourID --lr 1e-3 --interval 7 --adv_epochs 10  --batch_size 128 --adv_lr 0.01 --embed_size 30  

Yahoo:

python run_AdvDrop.py --modeltype AdvDrop --dataset Yahoo.new --n_layers 2 --neg_sample 1 --saveID yourID --lr 3e-3 --interval 15 --adv_epochs 5  --batch_size 128 --adv_lr 0.001 --embed_size 30  

KuaiRec:

python run_AdvDrop.py --modeltype AdvDrop --dataset KuaiRec.new --n_layers 2 --neg_sample 1 --saveID yourID --lr 5e-4 --interval 3 --adv_epochs 5  --batch_size 512 --adv_lr 0.001 --embed_size 30  

Yelp2018:

python run_AdvDrop.py --modeltype AdvDrop --dataset Yelp2018.new --n_layers 2 --neg_sample 1 --saveID yourID --lr 5e-4 --interval 7 --adv_epochs 15  --batch_size 1024 --adv_lr 0.01 --embed_size 64

• Douban:

python run_AdvDrop.py --modeltype AdvDrop --dataset Douban.new --n_layers 2 --neg_sample 1 --saveID yourID --lr 5e-4 --interval 10 --adv_epochs 3  --batch_size 4096 --adv_lr 0.01 --embed_size 64

CFC Training

python run_CFC.py --dataset Dataset --n_layers 2 --neg_sample 1 --saveID yourID --sample_mask 1 --embed_size embed --lr lr

Please replace Dataset with the name of the dataset you intend to evaluate and change the embedding size and learning rate as desired.

InvPref Training

Please see here InvPref.

AutoDebias Training

Please see here AutoDebias.

Other Baselines Training

python main.py --modeltype Model --dataset Dataset --n_layers 2 --neg_sample 1 --saveID yourID 

Please replace Model with the baseline name and replace Dataset with the name of the dataset you intend to evaluate. And don't forget to add specific params for the baselines.

Requirements

• python == 3.7.10

• pytorch == 1.12.1+cu102

• tensorflow == 1.14

• reckit == 0.2.4

• numpy == 1.20.1

• matplotlib == 3.3.4

• tqdm == 4.59.0

• scipy == 1.6.2

• pandas == 1.2.4

• cython ==0.29.23

Reference

If you want to use our codes and datasets in your research, please cite:

@inproceedings{AdvDrop,
  title={General Debiasing for Graph-based Collaborative Filtering via
Adversarial Graph Dropout},
  author={Zhang, An and Ma, Wenchang, and Wei, Pengbo, and Sheng, Leheng and Wang, Xiang},
  booktitle={{WWW}},
  year={2024}
}