GENET: Unleashing the Power of Side Information for Recommendation via Hypergraph Pretraining
.
├── config
│ ├── finetune.json
│ ├── finetune_seq.json
│ └── pretrain.json
── data
│ ├── data.zip
├── finetune
│ ├── finetuning_fewshots.py
│ ├── finetuning_gnn_signal.py
│ ├── finetuning_item_coldstart.py
│ ├── finetuning_user_coldstart.py
│ └── no_tuning.py
├── finetune_seq
│ └── finetuning_seq_signal.py
├── models
│ ├── HGNNP.py
│ └── Signal.py
├── pretrain
│ └── pretrain_lp_pm_plus.py
├── readme.md
├── save
│ ├── model
│ │ └── amazon
│ │ └── HGNNP_plus_LP_PM_500.pth
│ └── structure
│ └── amazon
│ └── hg_plus.pkl
└── utils
├── amazon.py
├── batch_test.py
├── cold_start.py
├── foursquare.py
├── gowalla.py
├── metrics.py
└── visualize.py
pip install -r requirements.txt # Install requirements with pip
We have provided the preprocessed Books datasets in the data/ folder. Please unzip the amazon.zip file before running the code. We have also provided the pre-trained model in the save/model/ folder. Please unzip them.
python finetune/finetuning_gnn_signal.py # Run a experiment, before that, please setup the root path of the dataset in the code.