How ouerhani-syrine/recommendation-system works
This is a small, runnable LightGCN recommendation demo. Running main.py displays a hard-coded user/item interaction dataset, then computes two rounds of embedding propagation from its normalized interaction graph and initial random embeddings. The code does not train a model, score items, or produce recommendations. The dataset and propagation implementations are excerpted; the graph shows only relationships visible in those excerpts.
Main entry points: main.py Demo script · dataset.py User-item data · lightgcn.py Layer propagation
Components 5
Runtime
- Demo script
main.py
Data preparation
- User-item data
src/dataset.py - Graph and embeddings
src/dataset.py
Embedding propagation
- Layer propagation
src/lightgcn.py
Other components
- Terminal output
Connections 5
- Demo script → User-item data: imports data
main.py - User-item data → Graph and embeddings: builds graph
src/dataset.py - Demo script → Terminal output: prints data
main.py - Demo script → Layer propagation: calls propagation
main.py - Demo script → Graph and embeddings: passes inputs
main.py