CF-KAN is a recommendation system (collaborative filtering) based on the Kolmogorov-Arnold Network (KAN, https://arxiv.org/abs/2404.19756) approach. This project leverages and explores the power of KAN for CF-based recommendation.
(For KAN implementation, I've refered to this repo: https://github.com/Blealtan/efficient-kan)
To get started with CF-KAN, clone the repository and install the required dependencies (pytorch and scipy, etc)
If you are ready, test it with the notebook file (CF-KAN.ipynb)
Dataset : ML-1M
We found only 10 epochs training on MovieLens-1M produces pretty convincing results!
- Recall@20: 0.2013
- NDCG@20: 0.1315
!! We will keep update this project! open to new collaboration :)
!! If you found this project is helpful for your research, please STAR this repository