sunshinelium/graph-icf

Graph-ICF: Item-based Collaborative Filtering based on Graph Neural Network. Knowledge-based systerms

Enviroment Requirement

pip install -r requirements.txt

• change base directory

Change ROOT_PATH in code/world.py

• command/model中包含三个模型，lgn（LightGCN),icf(Graph-ICF),graph-icf(尝试将icf与lgn结合起来的版本，未完成). 因此icf_ratio在模型lgn与icf下无效。

cd code && CUDA_VISIBLE_DEVICES=3 python main.py --decay=1e-4 --lr=0.001 --layer=3 --seed=2020 --dataset="ml-1m" --topks="[10, 20]" --recdim=64 --icf_ratio=0.5 --model=icf --comment=''

Update

1. Change the print format of each epoch
2. Add Cpp Extension in code/sources/ for negative sampling. To use the extension, please install pybind11 and cppimport under your environment