This is our implementation of ENMF: Efficient Neural Matrix Factorization (TOIS. 38, 2020). This also provides a fair evaluation of existing state-of-the-art recommendation models.

The code repository for the paper: Peijie et al., Neighborhood-Enhanced Supervised Contrastive Learning for Collaborative Filtering. IEEE TKDE, 2023.

用pytorch 方法复现了二十多个经典的推荐算法论文，其中包含排序论文和推荐召回论文，并在demo里面选了一个召回模型和排序模型的运行示例。

Source code for the RecSys 2024 paper "Revisiting LightGCN: Unexpected Inflexibility, Inconsistency, and A Remedy Towards Improved Recommendation."

AI Recommendation System — Real-time recommendation engine for movies, video, and digital media powered by SASRec Transformers, LightGCN Graphs, and PySpark Delta Lake.

Source code for the ICML 2026 paper "Rethinking Contrastive Learning in Graph Collaborative Filtering: Limitations and A Simple Remedy"

Deep learning music matcher that connects users based on complex listening patterns using GNN embeddings and Last.fm data.

A movie recommendation system utilizing a Graph Neural Network (GNN) framework implemented in Jupyter Notebook

rusket 🦀🧺

An institutional Deal Flow OS connecting founders and investors. Built on a Next.js/FastAPI microservices architecture, it features a PyTorch LightGCN/SBERT recommendation engine, a Zero-Knowledge E2EE Vault for secure data, and Groq-powered GenAI for automated due diligence.

RecSys Codes based on PyTorch

PyTorch implementation of LightGCN (He et al., SIGIR 2020) trained on Gowalla - achieves 97.3% of paper's Recall@20 in 400 epochs.

Comparative study of graph-based recommendation models NGCF and LightGCN, reproducing benchmark results on Amazon-book and evaluating generalization on Amazon-software-2023 using Recall@20 and NDCG@20.

Implementation of various collaborative filtering methods for recommender systems with implicit feedback

Heterogeneous graph neural network music recommender with Playlist-Track-Artist relations. Extends Stanford CS224W article with 5 GNN architectures.

A Recommender System for Google Maps reviews using LightGCN.

🧴 妆策AI 面向美妆新零售实战场景的智能推荐与营销转化平台 | AI-powered Beauty New Retail Platform | 15+ Algorithms | GBDT AUC=0.9993 | LightGCN HR@10=0.8413 | Six-Dimension Scoring

Graph-based movie recommendation system using ItemKNN, MF-BPR, LightGCN, and GraphSAGE on the MovieLens 20M dataset.

GNN-based recommendation under extreme data sparsity for WuxiaWorld web serials

Here is my implementation of a recommendation system using gated lightgcn and transact.