Paragon

Paragon is an innovative project that combines recommendation systems with diffusion models.

News

Accepted by Recsys 2025！

Project Structure

The project is organized into two main components:

1. Recommendation Model (Rec folder)

   • Training and testing code for the recommendation model
   • Located in the Rec folder

2. Diffusion Model (DIFFUSION folder)

   • Training code for the diffusion model
   • Code to generate parameters for the recommendation model
   • Located in the DIFFUSION folder

Installation

To set up the project, follow these steps:

git clone https:*******************/Paragon.git
cd Paragon
pip install -r requirements.txt

Usage

Recommendation Model

To train the recommendation model:

cd Rec
# Then find the config.ini file and change the mode inside it to train
python Rec_mian.py

To test a trained or generated recommendation model:

cd Rec
# Then find the config.ini file and change the mode inside it to test
python test_Recmodels.py

Diffusion Model

To train the diffusion model and generate parameters for the recommendation model:

cd DIFFUSION
python main.py

Citation

@article{shen2024generating,
title={Generating Model Parameters for Controlling: Parameter Diffusion for Controllable Multi-Task Recommendation},
author={Shen, Chenglei and Zhao, Jiahao and Zhang, Xiao and Yu, Weijie and He, Ming and Fan, Jianping},
journal={arXiv preprint arXiv:2410.10639},
year={2024}
}

Contributing

We welcome contributions to Paragon! Please feel free to submit pull requests or open issues to suggest improvements or report bugs.