Skip to content

This is the official repository of "EDGE-Rec: Efficient and Data-Guided Edge Diffusion for Recommender System Graphs" (GenAIRecP @ KDD 2024)

Notifications You must be signed in to change notification settings

EdoardoBotta/EDGE-Rec

 
 

Repository files navigation

EDGE-Rec: Efficient and Data-Guided Edge Diffusion for Recommender System Graphs

Authors: Edoardo Botta, Utkarsh Priyam, Hemit Shah

This is the official repository of "EDGE-Rec: Efficient and Data-Guided Edge Diffusion for Recommender System Graphs", submitted as final project for the CMU course 10-708: Probabilistic Graphical Models.

Training and Evaluation

We train and evaluate the model on the ML-100k dataset. We construct a custom 90-10 train-test split of the edges by adopting a stratified sampling approach to ensure that each user is represented in both the training and validation split.

We train on 1000 diffusion steps for 10000 iterations on a single A100 GPU in the Google Colab environment with batch size 16, patch size 50.

Replicability

Results can be replicated in a step-by-step fashion by running the execute.ipynb notebook.

Acknowledgments

The denoising diffusion model borrows from denoising-diffusion-pytorch.

About

This is the official repository of "EDGE-Rec: Efficient and Data-Guided Edge Diffusion for Recommender System Graphs" (GenAIRecP @ KDD 2024)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 88.6%
  • Jupyter Notebook 11.4%