Skip to content

Latest commit

 

History

History
74 lines (49 loc) · 2.89 KB

README.md

File metadata and controls

74 lines (49 loc) · 2.89 KB
Raw

FairSync: Ensuring Amortized Group Exposure in Distributed Recommendation Retrieval of WWW'24

Xu Chen, Ph.D. student of Renming University of China, GSAI

Any question, please mail to xc_chen@ruc.edu.cn

Prerequisites

  • Python 3
  • TensorFlow-GPU >= 1.8 (< 2.0)
  • pytorch
  • Faiss-GPU
  • The project is based on ComiRec github

Getting Started

Installation

Dataset

To fit the requirement of uploading, we only save the processed category files, for the processed dataset, please see ComiRec(KDD2020) https://github.com/THUDM/ComiRec

Training

Firstly, Training on the base retrieval model based on the following command

You can use python src/train.py --dataset {dataset_name} --model_type {model_name} to train a specific model on a dataset. Other hyperparameters can be found in the code. (If you share the server with others or you want to use the specific GPU(s), you may need to set CUDA_VISIBLE_DEVICES.)

For example, you can use python src/train.py --dataset book --model_type ComiRec-DR to train ComiRec-DR model on Book dataset.

Secondly, Process FairSync:

please run python src/FairSync.py

demo result: {'recall': 0.074, 'ndcg': 0.062, 'hitrate': 0.160, 'diversity': 0.208, 'ESP': 1.0}

For other parameters see:

args_name type description
dataset str Choose dataset in ["book", "taobao"]
model_type str The base model type, please make sure the model is trained in the first step
minimum_exposure float Each groups' required minimum exposures
topN int Retreival number, please make sure the model is trained in the first step
FairSync_lr float The dual updating learning rate
eval_batch_size int The performing batch size B in the paper

##For citation, please cite the following bib

@inproceedings{Xu-FairSync-WWW24,
author = {Xu, Chen and Jun Xu and Yiming Ding and Xiao Zhang and Qi Qi},
title = {FairSync: Ensuring Amortized Group Exposure in Distributed Recommendation Retrieval},
year = {2024},
isbn = {979-8-4007-0171-9/24/05},
publisher = {Association for Computing Machinery},
address = {Singapore, Singapore},
doi = {10.1145/3589334.3645413},
booktitle = {Proceedings of the ACM Web Conference 2024 (WWW '24)},
series = {WWW '24}
}