Create a recomendation system solution for MovieLens dataset. The dataset is in the directory.
Two models are presented in the directory: One was trained in ipynb file (LightGCN) LayerGCN was trained with LayerGCN main.
The result for both models are well-described in report:
- LightGCN:
Model weights are stored in lightgcn.pth.
The LightGCN notebook solution is based on Recommender Systems with GNNs in PyG.
- LayerGCN:
Solution is based on GitHub, the repo structure is was too complicated to split the solution into .py and .ipynb files, therefore I modified the repo to create plots and will make pull request. For now, the modified repo version is in the folder
There is no well-performed model, therefore all pretrained models are saved
_0.1225.png)
_0.1225.png)